Publications

@ARTICLE{10436711,
  author={Girgis, Antonious M. and Diggavi, Suhas},
  journal={IEEE Journal on Selected Areas in Information Theory}, 
  title={Multi-Message Shuffled Privacy in Federated Learning}, 
  year={2024},
  volume={5},
  number={},
  pages={12-27},
  abstract={We study the distributed mean estimation (DME) problem under privacy and communication constraints in the local differential privacy (LDP) and multi-message shuffled (MMS) privacy frameworks. The DME has wide applications in both federated learning and analytics. We propose a communication-efficient and differentially private algorithm for DME of bounded  $\ell _{2}$ -norm and  $\ell _{\infty }$ -norm vectors. We analyze our proposed DME schemes showing that our algorithms have order-optimal privacy-communication-performance trade-offs. Our algorithms are designed by giving unequal privacy assignments at different resolutions of the vector (through binary expansion) and appropriately combining it with coordinate sampling. These results are directly applied to give guarantees on private federated learning algorithms. We also numerically evaluate the performance of our private DME algorithms.},
  keywords={Privacy;Servers;Differential privacy;Federated learning;Numerical models;Optimization;Estimation;Differential privacy;mean estimation;federated learning;shuffled model},
  doi={10.1109/JSAIT.2024.3366225},
  ISSN={2641-8770},
  month={},
  type = {2},
  tags = {journal,PFL},
  url_arxiv          = {https://doi.org/10.48550/arXiv.2302.11152},
}

We study the distributed mean estimation (DME) problem under privacy and communication constraints in the local differential privacy (LDP) and multi-message shuffled (MMS) privacy frameworks. The DME has wide applications in both federated learning and analytics. We propose a communication-efficient and differentially private algorithm for DME of bounded $\ell _{2}$ -norm and $\ell _{∞ }$ -norm vectors. We analyze our proposed DME schemes showing that our algorithms have order-optimal privacy-communication-performance trade-offs. Our algorithms are designed by giving unequal privacy assignments at different resolutions of the vector (through binary expansion) and appropriately combining it with coordinate sampling. These results are directly applied to give guarantees on private federated learning algorithms. We also numerically evaluate the performance of our private DME algorithms.

@ARTICLE{10502314,
  author={Hanna, Osama and Girgis, Antonious M. and Fragouli, Christina and Diggavi, Suhas},
  journal={IEEE Journal on Selected Areas in Information Theory}, 
  title={Differentially Private Stochastic Linear Bandits: (Almost) for Free}, 
  year={2024},
  volume={5},
  number={},
  pages={135-147},
  abstract={In this paper, we propose differentially private algorithms for the problem of stochastic linear bandits in the central, local and shuffled models. In the central model, we achieve almost the same regret as the optimal non-private algorithms, which means we get privacy for free. In particular, we achieve a regret of  $\tilde {O}\left({\sqrt {T}+{}\frac {1}{\varepsilon }}\right)$  matching the known lower bound for private linear bandits, while the best previously known algorithm achieves  $\tilde {O}\left({{}\frac {1}{\varepsilon }\sqrt {T}}\right)$ . In the local case, we achieve a regret of  $\tilde {O}\left({{}\frac {1}{\varepsilon }{\sqrt {T}}}\right)$  which matches the non-private regret for constant  $\varepsilon $ , but suffers a regret penalty when  $\varepsilon $  is small. In the shuffled model, we also achieve regret of  $\tilde {O}\left({\sqrt {T}+{}\frac {1}{\varepsilon }}\right)$  while the best previously known algorithm suffers a regret of  $\tilde {O}\left({{}\frac {1}{\varepsilon }{T^{3/5}}}\right)$ . Our numerical evaluation validates our theoretical results. Our results generalize for contextual linear bandits with known context distributions.},
  keywords={Privacy;Servers;Stochastic processes;Differential privacy;Information theory;Context modeling;Sensors;Bandits;contextual bandits;differential privacy;linear bandits;differentially private bandits},
  doi={10.1109/JSAIT.2024.3389954},
  ISSN={2641-8770},
  month={},
  tags={journal,DML,PDL},
  type={2},
  url_arxiv={https://arxiv.org/abs/2207.03445},
  }

In this paper, we propose differentially private algorithms for the problem of stochastic linear bandits in the central, local and shuffled models. In the central model, we achieve almost the same regret as the optimal non-private algorithms, which means we get privacy for free. In particular, we achieve a regret of $̃{O}łeft({\sqrt {T}+{}\frac {1}{ɛ }}i̊ght)$ matching the known lower bound for private linear bandits, while the best previously known algorithm achieves $̃{O}łeft({{}\frac {1}{ɛ }\sqrt {T}}i̊ght)$ . In the local case, we achieve a regret of $̃{O}łeft({{}\frac {1}{ɛ }{\sqrt {T}}}i̊ght)$ which matches the non-private regret for constant $ɛ$ , but suffers a regret penalty when $ɛ$ is small. In the shuffled model, we also achieve regret of $̃{O}łeft({\sqrt {T}+{}\frac {1}{ɛ }}i̊ght)$ while the best previously known algorithm suffers a regret of $̃{O}łeft({{}\frac {1}{ɛ }{T^{3/5}}}i̊ght)$ . Our numerical evaluation validates our theoretical results. Our results generalize for contextual linear bandits with known context distributions.

@ARTICLE{9691792,  
author={Singh, Navjot and Data, Deepesh and George, Jemin and Diggavi, Suhas},  journal={IEEE Transactions on Automatic Control}, 
title={SPARQ-SGD: Event-Triggered and Compressed Communication in Decentralized Optimization}, 
journal={{IEEE} Trans. Autom. Control.},  
volume       = {68},
number       = {2},
pages        = {721--736},
year         = {2023},
url          = {https://doi.org/10.1109/TAC.2022.3145576},
doi          = {10.1109/TAC.2022.3145576},
abstract={In this paper, we propose and analyze SPARQ-SGD, a communication efficient algorithm for decentralized training of large-scale machine learning models over a graph with n nodes, where communication efficiency is achieved using compressed exchange of local model parameters among neighboring nodes, which is triggered only when an event (a locally computable condition) is satisfied. Specifically, in SPARQ-SGD, each node takes a fixed number of local gradient steps and then checks if the model parameters have significantly changed compared to its last update; only when the change is beyond a certain threshold (specified by a design criterion), it compresses its local model parameters using both quantization and sparsification and communicates them to its neighbors. We prove that SPARQ-SGD converges as O(1/nT) and O(1/sqrt(nT)) in the strongly-convex and non-convex settings, respectively, matching the convergence rates of plain decentralized SGD. This demonstrates that we get communication efficiency achieved by aggressive compression, local iterations, and event-triggered communication essentially for free.},  
keywords={},  
doi={10.1109/TAC.2022.3145576},  
ISSN={1558-2523},  
month={},
type={2},
tags={journal,DML,CEDL},
url_arxiv={https://arxiv.org/abs/1910.14280},
}

In this paper, we propose and analyze SPARQ-SGD, a communication efficient algorithm for decentralized training of large-scale machine learning models over a graph with n nodes, where communication efficiency is achieved using compressed exchange of local model parameters among neighboring nodes, which is triggered only when an event (a locally computable condition) is satisfied. Specifically, in SPARQ-SGD, each node takes a fixed number of local gradient steps and then checks if the model parameters have significantly changed compared to its last update; only when the change is beyond a certain threshold (specified by a design criterion), it compresses its local model parameters using both quantization and sparsification and communicates them to its neighbors. We prove that SPARQ-SGD converges as O(1/nT) and O(1/sqrt(nT)) in the strongly-convex and non-convex settings, respectively, matching the convergence rates of plain decentralized SGD. This demonstrates that we get communication efficiency achieved by aggressive compression, local iterations, and event-triggered communication essentially for free.

@ARTICLE{10057224,
  author={Nikolopoulos, Pavlos and Srinivasavaradhan, Sundara Rajan and Guo, Tao and Fragouli, Christina and Diggavi, Suhas N.},
  journal={IEEE Transactions on Information Theory}, 
  title={Community-Aware Group Testing}, 
  year={2023},
  volume={69},
  number={7},
  pages={4361-4383},
  abstract={Group testing is a technique that can reduce the number of tests needed to identify infected members in a population, by pooling together multiple diagnostic samples. Despite the variety and importance of prior results, traditional work on group testing has typically assumed independent infections. However, contagious diseases among humans, like SARS-CoV-2, have an important characteristic: infections are governed by community spread, and are therefore correlated. In this paper, we explore this observation and we argue that taking into account the community structure when testing can lead to significant savings in terms of the number of tests required to guarantee a given identification accuracy. To show that, we start with a simplistic (yet practical) infection model, where the entire population is organized in (possibly overlapping) communities and the infection probability of an individual depends on the communities (s)he participates in. Given this model, we compute new lower bounds on the number of tests for zero-error identification and design community-aware group testing algorithms that can be optimal under assumptions. Finally, we demonstrate significant benefits over traditional, community-agnostic group testing via simulations using both noiseless and noisy tests. Shorter versions of this article, which contained a subset of the material, were presented in the work by Nikolopoulos et al. (2021, 2021).},
  keywords={},
  doi={10.1109/TIT.2023.3250119},
  ISSN={1557-9654},
  month={July},
  type={2},
  tags={journal,PET},
}

Group testing is a technique that can reduce the number of tests needed to identify infected members in a population, by pooling together multiple diagnostic samples. Despite the variety and importance of prior results, traditional work on group testing has typically assumed independent infections. However, contagious diseases among humans, like SARS-CoV-2, have an important characteristic: infections are governed by community spread, and are therefore correlated. In this paper, we explore this observation and we argue that taking into account the community structure when testing can lead to significant savings in terms of the number of tests required to guarantee a given identification accuracy. To show that, we start with a simplistic (yet practical) infection model, where the entire population is organized in (possibly overlapping) communities and the infection probability of an individual depends on the communities (s)he participates in. Given this model, we compute new lower bounds on the number of tests for zero-error identification and design community-aware group testing algorithms that can be optimal under assumptions. Finally, we demonstrate significant benefits over traditional, community-agnostic group testing via simulations using both noiseless and noisy tests. Shorter versions of this article, which contained a subset of the material, were presented in the work by Nikolopoulos et al. (2021, 2021).

@ARTICLE{10075678,
  author={Cao, Xuanyu and Başar, Tamer and Diggavi, Suhas and Eldar, Yonina C. and Letaief, Khaled B. and Poor, H. Vincent and Zhang, Junshan},
  journal={IEEE Journal on Selected Areas in Communications}, 
  title={Guest Editorial Communication-Efficient Distributed Learning Over Networks}, 
  year={2023},
  volume={41},
  number={4},
  pages={845-850},
  abstract={Distributed machine learning is envisioned as the bedrock of future intelligent networks, where agents exchange information with each other to train models collaboratively without uploading data to a central processor. Despite its broad applicability, a downside of distributed learning is the need for iterative information exchange between agents, which may lead to high communication overhead unaffordable in many practical systems with limited communication resources. To resolve this communication bottleneck, we need to devise communication-efficient distributed learning algorithms and protocols that can reduce the communication cost and simultaneously achieve satisfactory learning/optimization performance. Accomplishing this goal necessitates synergistic techniques from a diverse set of fields, including optimization, machine learning, wireless communications, game theory, and network/graph theory. This Special Issue is dedicated to communication-efficient distributed learning from multiple perspectives, including fundamental theories, algorithm design and analysis, and practical considerations.},
  keywords={},
  doi={10.1109/JSAC.2023.3241848},
  ISSN={1558-0008},
  month={April},
  type={2},
  tags={journal,CEDL,DML},
}

Distributed machine learning is envisioned as the bedrock of future intelligent networks, where agents exchange information with each other to train models collaboratively without uploading data to a central processor. Despite its broad applicability, a downside of distributed learning is the need for iterative information exchange between agents, which may lead to high communication overhead unaffordable in many practical systems with limited communication resources. To resolve this communication bottleneck, we need to devise communication-efficient distributed learning algorithms and protocols that can reduce the communication cost and simultaneously achieve satisfactory learning/optimization performance. Accomplishing this goal necessitates synergistic techniques from a diverse set of fields, including optimization, machine learning, wireless communications, game theory, and network/graph theory. This Special Issue is dedicated to communication-efficient distributed learning from multiple perspectives, including fundamental theories, algorithm design and analysis, and practical considerations.

@article{joshi2023hqalign,
    author = {Joshi, Dhaivat and Diggavi, Suhas and Chaisson, Mark J P and Kannan, Sreeram},
    title = "{HQAlign: aligning nanopore reads for SV detection using current-level modeling}",
    journal = {Bioinformatics},
    volume = {39},
    number = {10},
    year = {2023},
    month = {September},
    abstract = "{Detection of structural variants (SVs) from the alignment of sample DNA reads to the reference genome is an important problem in understanding human diseases. Long reads that can span repeat regions, along with an accurate alignment of these long reads play an important role in identifying novel SVs. Long-read sequencers, such as nanopore sequencing, can address this problem by providing very long reads but with high error rates, making accurate alignment challenging. Many errors induced by nanopore sequencing have a bias because of the physics of the sequencing process and proper utilization of these error characteristics can play an important role in designing a robust aligner for SV detection problems. In this article, we design and evaluate HQAlign, an aligner for SV detection using nanopore sequenced reads. The key ideas of HQAlign include (i) using base-called nanopore reads along with the nanopore physics to improve alignments for SVs, (ii) incorporating SV-specific changes to the alignment pipeline, and (iii) adapting these into existing state-of-the-art long-read aligner pipeline, minimap2 (v2.24), for efficient alignments.We show that HQAlign captures about 4\\%–6\\% complementary SVs across different datasets, which are missed by minimap2 alignments while having a standalone performance at par with minimap2 for real nanopore reads data. For the common SV calls between HQAlign and minimap2, HQAlign improves the start and the end breakpoint accuracy by about 10\\%–50\\% for SVs across different datasets. Moreover, HQAlign improves the alignment rate to 89.35\\% from minimap2 85.64\\% for nanopore reads alignment to recent telomere-to-telomere CHM13 assembly, and it improves to 86.65\\% from 83.48\\% for nanopore reads alignment to GRCh37 human genome.https://github.com/joshidhaivat/HQAlign.git.}",
    issn = {1367-4811},
    doi = {10.1093/bioinformatics/btad580},
    url_arxiv={https://arxiv.org/abs/2301.03834},
    url_biorxiv={https://doi.org/10.1101/2023.01.08.523172},
    eprint = {https://academic.oup.com/bioinformatics/article-pdf/39/10/btad580/52147189/btad580.pdf},
    tags={journal,BioInf,NDS},
    type={2},
}

Detection of structural variants (SVs) from the alignment of sample DNA reads to the reference genome is an important problem in understanding human diseases. Long reads that can span repeat regions, along with an accurate alignment of these long reads play an important role in identifying novel SVs. Long-read sequencers, such as nanopore sequencing, can address this problem by providing very long reads but with high error rates, making accurate alignment challenging. Many errors induced by nanopore sequencing have a bias because of the physics of the sequencing process and proper utilization of these error characteristics can play an important role in designing a robust aligner for SV detection problems. In this article, we design and evaluate HQAlign, an aligner for SV detection using nanopore sequenced reads. The key ideas of HQAlign include (i) using base-called nanopore reads along with the nanopore physics to improve alignments for SVs, (ii) incorporating SV-specific changes to the alignment pipeline, and (iii) adapting these into existing state-of-the-art long-read aligner pipeline, minimap2 (v2.24), for efficient alignments.We show that HQAlign captures about 4\%–6\% complementary SVs across different datasets, which are missed by minimap2 alignments while having a standalone performance at par with minimap2 for real nanopore reads data. For the common SV calls between HQAlign and minimap2, HQAlign improves the start and the end breakpoint accuracy by about 10\%–50\% for SVs across different datasets. Moreover, HQAlign improves the alignment rate to 89.35\% from minimap2 85.64\% for nanopore reads alignment to recent telomere-to-telomere CHM13 assembly, and it improves to 86.65\% from 83.48\% for nanopore reads alignment to GRCh37 human genome.https://github.com/joshidhaivat/HQAlign.git.

@article{DBLP:journals/jsac/CaoBDELPZ23a,
  author       = {Xuanyu Cao and
                  Tamer Basar and
                  Suhas N. Diggavi and
                  Yonina C. Eldar and
                  Khaled B. Letaief and
                  H. Vincent Poor and
                  Junshan Zhang},
  title        = {Communication-Efficient Distributed Learning: An Overview},
  journal      = {{IEEE} J. Sel. Areas Commun.},
  volume       = {41},
  number       = {4},
  pages        = {851--873},
  year         = {2023},
  url          = {https://doi.org/10.1109/JSAC.2023.3242710},
  doi          = {10.1109/JSAC.2023.3242710},
  timestamp    = {Tue, 28 Mar 2023 19:50:24 +0200},
  biburl       = {https://dblp.org/rec/journals/jsac/CaoBDELPZ23a.bib},
  type    = {2},
  tags = {journal,CEDL,DML}
}

@article{ozkara2022generative,
  title={A Generative Framework for Personalized Learning and Estimation: Theory, Algorithms, and Privacy},
  author={Ozkara, Kaan and Girgis, Antonious M and Data, Deepesh and Diggavi, Suhas},
  journal={in International Conference on Learning Representations (ICLR)},
  year={2023},
  tags={conf,PFL},
  type={4},
  url_arxiv={https://arxiv.org/abs/2207.01771},
  url={https://openreview.net/forum?id=FUiDMCr_W4o},
  abstract={A distinguishing characteristic of federated learning is that the (local) client data could have statistical heterogeneity. This heterogeneity has motivated the design of personalized learning, where individual (personalized) models are trained, through collaboration. There have been various personalization methods proposed in literature, with seemingly very different forms and methods ranging from use of a single global model for local regularization and model interpolation, to use of multiple global models for personalized clustering, etc. In this work, we begin with a generative framework that could potentially unify several different algorithms as well as suggest new algorithms. We apply our generative framework to personalized estimation, and connect it to the classical empirical Bayes' methodology. We develop private personalized estimation under this framework. We then use our generative framework for learning, which unifies several known personalized FL algorithms and also suggests new ones; we propose and study a new algorithm AdaPeD based on a Knowledge Distillation, which numerically outperforms several known algorithms. We also develop privacy for personalized learning methods with guarantees for user-level privacy and composition. We numerically evaluate the performance as well as the privacy for both the estimation and learning problems, demonstrating the advantages of our proposed methods.},
}

A distinguishing characteristic of federated learning is that the (local) client data could have statistical heterogeneity. This heterogeneity has motivated the design of personalized learning, where individual (personalized) models are trained, through collaboration. There have been various personalization methods proposed in literature, with seemingly very different forms and methods ranging from use of a single global model for local regularization and model interpolation, to use of multiple global models for personalized clustering, etc. In this work, we begin with a generative framework that could potentially unify several different algorithms as well as suggest new algorithms. We apply our generative framework to personalized estimation, and connect it to the classical empirical Bayes' methodology. We develop private personalized estimation under this framework. We then use our generative framework for learning, which unifies several known personalized FL algorithms and also suggests new ones; we propose and study a new algorithm AdaPeD based on a Knowledge Distillation, which numerically outperforms several known algorithms. We also develop privacy for personalized learning methods with guarantees for user-level privacy and composition. We numerically evaluate the performance as well as the privacy for both the estimation and learning problems, demonstrating the advantages of our proposed methods.

@ARTICLE{9725266,
  author={Sebastian, Joyson and Diggavi, Suhas},
  journal={IEEE Transactions on Wireless Communications}, 
  title={On the Generalized Degrees of Freedom of the Noncoherent Interference Channel}, 
  year={2022},
  volume={21},
  number={9},
  pages={7011-7025},
  abstract={We study the generalized degrees of freedom (gDoF) of the block-fading noncoherent 2-user interference channel (IC) with a coherence time of  $T$  symbol durations and symmetric fading statistics. We demonstrate that a standard training-based scheme for the noncoherent IC is suboptimal in several regimes. We study and analyze several alternate schemes: the first is a new noncoherent scheme using rate-splitting. We also consider a scheme that treats interference-as-noise (TIN) and a time division multiplexing (TDM) scheme. We show that a standard training-based scheme for the noncoherent IC is outperformed by one of these schemes in several regimes: our results demonstrate that in the very weak interference regime, the TIN scheme is the best; in the strong interference regime, the TDM scheme and the noncoherent rate-splitting scheme give better performance; in other cases either of the TIN, TDM or noncoherent rate-splitting scheme could be preferred. We also study the noncoherent IC with feedback and propose another noncoherent rate-splitting scheme. Again for the feedback case, our results demonstrate that a standard training-based scheme can be outperformed by other schemes.},
  keywords={},
  doi={10.1109/TWC.2022.3153893},
  ISSN={1558-2248},
  month={Sep.},
  tags={journal,WiNetnew},
  type={2},
  url_arxiv={https://arxiv.org/abs/1812.03579},
}

We study the generalized degrees of freedom (gDoF) of the block-fading noncoherent 2-user interference channel (IC) with a coherence time of $T$ symbol durations and symmetric fading statistics. We demonstrate that a standard training-based scheme for the noncoherent IC is suboptimal in several regimes. We study and analyze several alternate schemes: the first is a new noncoherent scheme using rate-splitting. We also consider a scheme that treats interference-as-noise (TIN) and a time division multiplexing (TDM) scheme. We show that a standard training-based scheme for the noncoherent IC is outperformed by one of these schemes in several regimes: our results demonstrate that in the very weak interference regime, the TIN scheme is the best; in the strong interference regime, the TDM scheme and the noncoherent rate-splitting scheme give better performance; in other cases either of the TIN, TDM or noncoherent rate-splitting scheme could be preferred. We also study the noncoherent IC with feedback and propose another noncoherent rate-splitting scheme. Again for the feedback case, our results demonstrate that a standard training-based scheme can be outperformed by other schemes.

@ARTICLE{comm-efficientDL,
  author={Can, S. and Basar, T. and Diggavi, S. N. and Eldar, Y. C. and Letaief K. B. and Poor, H. V. and Zhang, J.},
  journal={to appear in IEEE Journal of Selected Areas in Communications (JSAC), 2022.}, 
  title={Communication-Efficient Distributed Learning: An Overview}, 
  year={2022},
  keywords={},
  doi={},
  ISSN={},
  month={},
  tags={journal,CEDL,DML},
  type={2},
  }

@INPROCEEDINGS{9834713,
  author={Girgis, Antonious M. and Data, Deepesh and Diggavi, Suhas},
  booktitle={IEEE International Symposium on Information Theory (ISIT)}, 
  title={Distributed User-Level Private Mean Estimation}, 
  year={2022},
  volume={},
  number={},
  pages={2196-2201},
  doi={10.1109/ISIT50566.2022.9834713},
  type={4},
  tags={conf,DML,PDL},
}

@INPROCEEDINGS{9834736,
  author={Rammal, Mohamad Rida and Diggavi, Suhas and Sabharwal, Ashutosh},
  booktitle={IEEE International Symposium on Information Theory (ISIT)}, 
  title={3D Orientation Estimation With Configurable Backscatter Arrays}, 
  year={2022},
  volume={},
  number={},
  pages={1832-1837},
  abstract={We consider the problem of estimating the orientation of a 3D object with the assistance of configurable backscatter tags. We explore the idea of designing tag response codes to improve the accuracy of orientation estimation. To minimize the difference between the true and estimated orientation, we propose two code design criteria. We also derive a lower bound on the worst-case error using Le Cam’s method and provide simulation results for multiple scenarios including perfect and imperfect channel knowledge, comparing the performance of various coding methods against the suggested designs.},
  keywords={},
  doi={10.1109/ISIT50566.2022.9834736},
  ISSN={2157-8117},
  month={June},
  tags={conf,WiNetnew,WiIm},
  type={4},
}

We consider the problem of estimating the orientation of a 3D object with the assistance of configurable backscatter tags. We explore the idea of designing tag response codes to improve the accuracy of orientation estimation. To minimize the difference between the true and estimated orientation, we propose two code design criteria. We also derive a lower bound on the worst-case error using Le Cam’s method and provide simulation results for multiple scenarios including perfect and imperfect channel knowledge, comparing the performance of various coding methods against the suggested designs.

@INPROCEEDINGS{9834790,
  author={Hanna, Osama A. and Li, Xinlin and Fragouli, Christina and Diggavi, Suhas},
  booktitle={IEEE International Symposium on Information Theory (ISIT)}, 
  title={Can we break the dependency in distributed detection?}, 
  year={2022},
  volume={},
  number={},
  pages={2720-2725},
  abstract={We consider a distributed detection problem where sensors observe dependent observations. We ask, if we can allow the sensors to locally exchange a few bits with each other, whether we can use these bits to "break" the dependency of the sensor observations, and thus reduce the dependent detection problem to the much better-studied and understood case of conditionally independent observations. To this end, we propose an optimization problem that we prove is equivalent to minimizing the dependency between the sensor observations. This problem is in general NP-hard, however, we show that for at least some cases of Gaussian distributions it can be solved efficiently. For general distributions, we propose to use alternating minimization and derive a constant factor approximation algorithm. Numerical evaluations indicate that our approach can offer significant improvement in detection accuracy over alternative schemes.},
  keywords={},
  doi={10.1109/ISIT50566.2022.9834790},
  ISSN={2157-8117},
  month={June},
  tags={conf,CEDL,DML},
  type={4},
}

We consider a distributed detection problem where sensors observe dependent observations. We ask, if we can allow the sensors to locally exchange a few bits with each other, whether we can use these bits to "break" the dependency of the sensor observations, and thus reduce the dependent detection problem to the much better-studied and understood case of conditionally independent observations. To this end, we propose an optimization problem that we prove is equivalent to minimizing the dependency between the sensor observations. This problem is in general NP-hard, however, we show that for at least some cases of Gaussian distributions it can be solved efficiently. For general distributions, we propose to use alternating minimization and derive a constant factor approximation algorithm. Numerical evaluations indicate that our approach can offer significant improvement in detection accuracy over alternative schemes.

@INPROCEEDINGS{9834823,
  author={Srinivasavaradhan, Sundara Rajan and Nikolopoulos, Pavlos and Fragouli, Christina and Diggavi, Suhas},
  booktitle={2022 IEEE International Symposium on Information Theory (ISIT)}, 
  title={Dynamic group testing to control and monitor disease progression in a population}, 
  year={2022},
  volume={},
  number={},
  pages={2255-2260},
  abstract={In this paper, we introduce a "discrete-time SIR stochastic block model" that also allows for group testing and interventions on a daily basis. Our model can be regarded as a discrete version of the well-known continuous-time SIR stochastic network model [1] and relies on a specific type of weighted graph to capture the underlying community spread. Given that infection model, we then formulate a dynamic group-testing problem by asking: (a) what is the minimum number of tests needed everyday to identify all infections? and (b) are there nonadaptive group testing strategies that achieve this with vanishing error probability? Our results show that one can leverage the knowledge of the community infection model to compute a lower bound on the number of tests and also inform nonadaptive group testing algorithms, so that they can achieve (almost) the same performance as complete individual testing with a much smaller number of tests. Moreover, these algorithms are order-optimal, under specific conditions.},
  keywords={},
  doi={10.1109/ISIT50566.2022.9834823},
  ISSN={2157-8117},
  month={June},
  tags={conf,PET},
  type={4},
  url_arxiv={https://arxiv.org/abs/2106.10765},
}

In this paper, we introduce a "discrete-time SIR stochastic block model" that also allows for group testing and interventions on a daily basis. Our model can be regarded as a discrete version of the well-known continuous-time SIR stochastic network model [1] and relies on a specific type of weighted graph to capture the underlying community spread. Given that infection model, we then formulate a dynamic group-testing problem by asking: (a) what is the minimum number of tests needed everyday to identify all infections? and (b) are there nonadaptive group testing strategies that achieve this with vanishing error probability? Our results show that one can leverage the knowledge of the community infection model to compute a lower bound on the number of tests and also inform nonadaptive group testing algorithms, so that they can achieve (almost) the same performance as complete individual testing with a much smaller number of tests. Moreover, these algorithms are order-optimal, under specific conditions.

@INPROCEEDINGS{rammal2022leave,
  title={On Leave-One-Out Conditional Mutual Information For Generalization},
  author={Rammal, Mohamad Rida and Achille, Alessandro and Golatkar, Aditya and Diggavi, Suhas and Soatto, Stefano},
  booktitle={Advances in Neural Information Processing Systems (NeurIPS)},
  year={2022},
  type={4},
  tags={conf,MLIT,IT},
  url_arxiv={https://arxiv.org/abs/2207.00581},
}

@INPROCEEDINGS{mao2022decentralizedlearning,
  title={Decentralized Learning Robust to Data Poisoning Attacks},
  author={Mao, Y. and Data, D. and Diggavi, S. N. and Tabuada, P.},
  booktitle={IEEE Control and Decision Conference (CDC)},
  year={2022},
  type={4},
  tags={conf,SDL,DML},
}

@article{srinivasavaradhan2022improving,
  title={Improving Group Testing via Gradient Descent},
  author={Srinivasavaradhan, Sundara Rajan and Nikolopoulos, Pavlos and Fragouli, Christina and Diggavi, Suhas},
  journal={IEEE International Symposium on Information theory},
  year={2022},
  type={4},
  tags={conf,PET},
  pages={2243-2248},
  url_arxiv={https://arxiv.org/abs/2201.12325},
  abstract={We study the problem of group testing with non-identical, independent priors. So far, the pooling strategies that have been proposed in the literature take the following approach: a hand-crafted test design along with a decoding strategy is proposed, and guarantees are provided on how many tests are sufficient in order to identify all infections in a population. In this paper, we take a different, yet perhaps more practical, approach: we fix the decoder and the number of tests, and we ask, given these, what is the best test design one could use? We explore this question for the Definite Non-Defectives (DND) decoder. We formulate a (non-convex) optimization problem, where the objective function is the expected number of errors for a particular design. We find approximate solutions via gradient descent, which we further optimize with informed initialization. We illustrate through simulations that our method can achieve significant performance improvement over traditional approaches.},
}

We study the problem of group testing with non-identical, independent priors. So far, the pooling strategies that have been proposed in the literature take the following approach: a hand-crafted test design along with a decoding strategy is proposed, and guarantees are provided on how many tests are sufficient in order to identify all infections in a population. In this paper, we take a different, yet perhaps more practical, approach: we fix the decoder and the number of tests, and we ask, given these, what is the best test design one could use? We explore this question for the Definite Non-Defectives (DND) decoder. We formulate a (non-convex) optimization problem, where the objective function is the expected number of errors for a particular design. We find approximate solutions via gradient descent, which we further optimize with informed initialization. We illustrate through simulations that our method can achieve significant performance improvement over traditional approaches.

@article{singh2021decentralized,
  title={Decentralized Multi-Task Stochastic Optimization With Compressed Communications},
  author={Singh, Navjot and Cao, Xuanyu and Diggavi, Suhas and Basar, Tamer},
  journal={arXiv preprint arXiv:2112.12373},
  year={2021},
  type={1},
  tags={journalSub,CEDL,DML},
  url_arxiv={https://arxiv.org/abs/2112.12373},
  abstract={We consider a multi-agent network where each node has a stochastic (local) cost function that depends on the decision variable of that node and a random variable, and further the decision variables of neighboring nodes are pairwise constrained. There is an aggregate objective function for the network, composed additively of the expected values of the local cost functions at the nodes, and the overall goal of the network is to obtain the minimizing solution to this aggregate objective function subject to all the pairwise constraints. This is to be achieved at the node level using decentralized information and local computation, with exchanges of only compressed information allowed by neighboring nodes. The paper develops algorithms and obtains performance bounds for two different models of local information availability at the nodes: (i) sample feedback, where each node has direct access to samples of the local random variable to evaluate its local cost, and (ii) bandit feedback, where samples of the random variables are not available, but only the values of the local cost functions at two random points close to the decision are available to each node. For both models, with compressed communication between neighbors, we have developed decentralized saddle-point algorithms that deliver performances no different (in order sense) from those without communication compression; specifically, we show that deviation from the global minimum value and violations of the constraints are upper-bounded by (T−12) and (T−14), respectively, where T is the number of iterations. Numerical examples provided in the paper corroborate these bounds and demonstrate the communication efficiency of the proposed method.},
}

We consider a multi-agent network where each node has a stochastic (local) cost function that depends on the decision variable of that node and a random variable, and further the decision variables of neighboring nodes are pairwise constrained. There is an aggregate objective function for the network, composed additively of the expected values of the local cost functions at the nodes, and the overall goal of the network is to obtain the minimizing solution to this aggregate objective function subject to all the pairwise constraints. This is to be achieved at the node level using decentralized information and local computation, with exchanges of only compressed information allowed by neighboring nodes. The paper develops algorithms and obtains performance bounds for two different models of local information availability at the nodes: (i) sample feedback, where each node has direct access to samples of the local random variable to evaluate its local cost, and (ii) bandit feedback, where samples of the random variables are not available, but only the values of the local cost functions at two random points close to the decision are available to each node. For both models, with compressed communication between neighbors, we have developed decentralized saddle-point algorithms that deliver performances no different (in order sense) from those without communication compression; specifically, we show that deviation from the global minimum value and violations of the constraints are upper-bounded by (T−12) and (T−14), respectively, where T is the number of iterations. Numerical examples provided in the paper corroborate these bounds and demonstrate the communication efficiency of the proposed method.

@article{rammal2021coded,
  title={Coded Estimation: Design of Backscatter Array Codes for 3D Orientation Estimation},
  author={Rammal, Mohamad Rida and Diggavi, Suhas and Sabharwal, Ashutosh},
  journal={arXiv preprint arXiv:2112.00883},
  year={2021},
  url_arxiv={https://arxiv.org/abs/2112.00883},
  tags={journalSub,WiNetnew,WiIm},
  type={1},
  abstract={We consider the problem of estimating the orientation of a 3D object with the assistance of configurable backscatter tags. We explore the idea of designing tag response codes to improve the accuracy of orientation estimation. To minimize the difference between the true and estimated orientation, we propose two code design criteria. We also derive a lower bound on the worst-case error using Le Cam's method and provide simulation results for multiple scenarios including line-of-sight only and multipath, comparing the theoretical bounds to those achieved by the designs.},
}

We consider the problem of estimating the orientation of a 3D object with the assistance of configurable backscatter tags. We explore the idea of designing tag response codes to improve the accuracy of orientation estimation. To minimize the difference between the true and estimated orientation, we propose two code design criteria. We also derive a lower bound on the worst-case error using Le Cam's method and provide simulation results for multiple scenarios including line-of-sight only and multipath, comparing the theoretical bounds to those achieved by the designs.

@article{ozkara2021qupel,
  title = {QuPeL: Quantized Personalization with Applications to Federated Learning},
  author = {Ozkara, Kaan and Singh, Navjot and Data, Deepesh and Diggavi, Suhas},
  journal = {arXiv preprint arXiv:2102.11786},
  year = {2021},
  abstract = {Traditionally, federated learning (FL) aims to train a single global model while collaboratively using multiple clients and a server. Two natural challenges that FL algorithms face are heterogeneity in data across clients and collaboration of clients with {\em diverse resources}. In this work, we introduce a \textit{quantized} and \textit{personalized} FL algorithm QuPeL that facilitates collective training with heterogeneous clients while respecting resource diversity. For personalization, we allow clients to learn \textit{compressed personalized models} with different quantization parameters depending on their resources. Towards this, first we propose an algorithm for learning quantized models through a relaxed optimization problem, where quantization values are also optimized over. When each client participating in the (federated) learning process has different requirements of the quantized model (both in value and precision), we formulate a quantized personalization framework by introducing a penalty term for local client objectives against a globally trained model to encourage collaboration. We develop an alternating proximal gradient update for solving this quantized personalization problem, and we analyze its convergence properties. Numerically, we show that optimizing over the quantization levels increases the performance and we validate that QuPeL outperforms both FedAvg and local training of clients in a heterogeneous setting.},
  tags = {journalSub,CEDL,DML},
  url_arxiv = {https://arxiv.org/abs/2102.11786},
  type = {1},
}

Traditionally, federated learning (FL) aims to train a single global model while collaboratively using multiple clients and a server. Two natural challenges that FL algorithms face are heterogeneity in data across clients and collaboration of clients with \em diverse resources. In this work, we introduce a quantized and personalized FL algorithm QuPeL that facilitates collective training with heterogeneous clients while respecting resource diversity. For personalization, we allow clients to learn compressed personalized models with different quantization parameters depending on their resources. Towards this, first we propose an algorithm for learning quantized models through a relaxed optimization problem, where quantization values are also optimized over. When each client participating in the (federated) learning process has different requirements of the quantized model (both in value and precision), we formulate a quantized personalization framework by introducing a penalty term for local client objectives against a globally trained model to encourage collaboration. We develop an alternating proximal gradient update for solving this quantized personalization problem, and we analyze its convergence properties. Numerically, we show that optimizing over the quantization levels increases the performance and we validate that QuPeL outperforms both FedAvg and local training of clients in a heterogeneous setting.

@article{9112310,
 abstract = {Secure state-reconstruction is the problem of reconstructing the state of a linear time-invariant system from sensor measurements that have been corrupted by an adversary. Whereas most work focuses on attacks on sensors, we consider the more challenging case where attacks occur on sensors as well as on nodes and links of a network that transports sensor measurements to a receiver. In this letter we provide necessary and sufficient conditions for the secure state-reconstruction problem to be solvable in the presence of attacks on sensors and on the network.},
 author = {Y. {Mao} and S. {Diggavi} and C. {Fragouli} and P. {Tabuada}},
 doi = {10.1109/LCSYS.2020.3000853},
 issn = {2475-1456},
 journal = {IEEE Control Systems Letters},
 keywords = {control system security;linear systems;secure state-reconstruction;sensor measurements;linear time-invariant system;Receivers;Sensors;Relay networks (telecommunications);Observability;Indexes;Network coding;Fault tolerant systems;network analysis and control;fault detection},
 month = {Jan},
 number = {1},
 pages = {157-162},
 tags = {journal,CPS},
 title = {Secure State-Reconstruction Over Networks Subject to Attacks},
 type = {2},
 volume = {5},
 year = {2021}
}

Secure state-reconstruction is the problem of reconstructing the state of a linear time-invariant system from sensor measurements that have been corrupted by an adversary. Whereas most work focuses on attacks on sensors, we consider the more challenging case where attacks occur on sensors as well as on nodes and links of a network that transports sensor measurements to a receiver. In this letter we provide necessary and sufficient conditions for the secure state-reconstruction problem to be solvable in the presence of attacks on sensors and on the network.

@ARTICLE{9344807,
  author={Girgis, Antonious M. and Data, Deepesh and Diggavi, Suhas and Kairouz, Peter and Suresh, Ananda Theertha},
  journal={IEEE Journal on Selected Areas in Information Theory}, 
  title={Shuffled Model of Federated Learning: Privacy, Accuracy and Communication Trade-Offs}, 
  year={2021},
  volume={2},
  number={1},
  pages={464-478},
  abstract={We consider a distributed empirical risk minimization (ERM) optimization problem with communication efficiency and privacy requirements, motivated by the federated learning (FL) framework (Kairouz et al., 2019). Unique challenges to the traditional ERM problem in the context of FL include  $\textsf{(i)}$  need to provide privacy guarantees on clients’ data,  $\textsf{(ii)}$  compress the communication between clients and the server, since clients might have low-bandwidth links,  $\textsf{(iii)}$  work with a dynamic client population at each round of communication between the server and the clients, as a small fraction of clients are sampled at each round. To address these challenges we develop (optimal) communication-efficient schemes for private mean estimation for several  $\ell _{p}$  spaces, enabling efficient gradient aggregation for each iteration of the optimization solution of the ERM. We also provide lower and upper bounds for mean estimation with privacy and communication constraints for arbitrary  $\ell _{p}$  spaces. To get the overall communication, privacy, and optimization performance operation point, we combine this with privacy amplification opportunities inherent to this setup. Our solution takes advantage of the inherent privacy amplification provided by client sampling and data sampling at each client (through Stochastic Gradient Descent) as well as the recently developed privacy framework using anonymization, which effectively presents to the server responses that are randomly shuffled with respect to the clients. Putting these together, we demonstrate that one can get the same privacy, optimization-performance operating point developed in recent methods that use full-precision communication, but at a much lower communication cost, i.e., effectively getting communication efficiency for “free”.},
  keywords={},
  doi={10.1109/JSAIT.2021.3056102},
  ISSN={2641-8770},
  month={March},
  type = {2},
  tags={journal,DML,PDL},
}

We consider a distributed empirical risk minimization (ERM) optimization problem with communication efficiency and privacy requirements, motivated by the federated learning (FL) framework (Kairouz et al., 2019). Unique challenges to the traditional ERM problem in the context of FL include $\textsf{(i)}$ need to provide privacy guarantees on clients’ data, $\textsf{(ii)}$ compress the communication between clients and the server, since clients might have low-bandwidth links, $\textsf{(iii)}$ work with a dynamic client population at each round of communication between the server and the clients, as a small fraction of clients are sampled at each round. To address these challenges we develop (optimal) communication-efficient schemes for private mean estimation for several $\ell _{p}$ spaces, enabling efficient gradient aggregation for each iteration of the optimization solution of the ERM. We also provide lower and upper bounds for mean estimation with privacy and communication constraints for arbitrary $\ell _{p}$ spaces. To get the overall communication, privacy, and optimization performance operation point, we combine this with privacy amplification opportunities inherent to this setup. Our solution takes advantage of the inherent privacy amplification provided by client sampling and data sampling at each client (through Stochastic Gradient Descent) as well as the recently developed privacy framework using anonymization, which effectively presents to the server responses that are randomly shuffled with respect to the clients. Putting these together, we demonstrate that one can get the same privacy, optimization-performance operating point developed in recent methods that use full-precision communication, but at a much lower communication cost, i.e., effectively getting communication efficiency for “free”.

@ARTICLE{9513259,
  author={Singh, Navjot and Data, Deepesh and George, Jemin and Diggavi, Suhas},
  journal={IEEE Journal on Selected Areas in Information Theory}, 
  title={SQuARM-SGD: Communication-Efficient Momentum SGD for Decentralized Optimization}, 
  year={2021},
  volume={2},
  number={3},
  pages={954-969},
  abstract={In this paper, we propose and analyze SQuARM-SGD, a communication-efficient algorithm for decentralized training of large-scale machine learning models over a network. In SQuARM-SGD, each node performs a fixed number of local SGD steps using Nesterov’s momentum and then sends sparsified and quantized updates to its neighbors regulated by a locally computable triggering criterion. We provide convergence guarantees of our algorithm for general (non-convex) and convex smooth objectives, which, to the best of our knowledge, is the first theoretical analysis for compressed decentralized SGD with momentum updates. We show that the convergence rate of SQuARM-SGD matches that of vanilla SGD. We empirically show that including momentum updates in SQuARM-SGD can lead to better test performance than the current state-of-the-art which does not consider momentum updates.},
  keywords={},
  doi={10.1109/JSAIT.2021.3103920},
  ISSN={2641-8770},
  month={Sep.},
  tags = {journal,CEDL,DML},
  type = {2},
  url_arxiv = {https://arxiv.org/abs/2005.07041},
  }

In this paper, we propose and analyze SQuARM-SGD, a communication-efficient algorithm for decentralized training of large-scale machine learning models over a network. In SQuARM-SGD, each node performs a fixed number of local SGD steps using Nesterov’s momentum and then sends sparsified and quantized updates to its neighbors regulated by a locally computable triggering criterion. We provide convergence guarantees of our algorithm for general (non-convex) and convex smooth objectives, which, to the best of our knowledge, is the first theoretical analysis for compressed decentralized SGD with momentum updates. We show that the convergence rate of SQuARM-SGD matches that of vanilla SGD. We empirically show that including momentum updates in SQuARM-SGD can lead to better test performance than the current state-of-the-art which does not consider momentum updates.

@ARTICLE{9521982,
  author={Hanna, Osama A. and Ezzeldin, Yahya H. and Fragouli, Christina and Diggavi, Suhas},
  journal={IEEE Journal on Selected Areas in Information Theory}, 
  title={Quantization of Distributed Data for Learning}, 
  year={2021},
  volume={2},
  number={3},
  pages={987-1001},
  abstract={We consider machine learning applications that train a model by leveraging data distributed over a trusted network, where communication constraints can create a performance bottleneck. A number of recent approaches propose to overcome this bottleneck through compression of gradient updates. However, as models become larger, so does the size of the gradient updates. In this paper, we propose an alternate approach to learn from distributed data that quantizes data instead of gradients, and can support learning over applications where the size of gradient updates is prohibitive. Our approach leverages the dependency of the computed gradient on data samples, which lie in a much smaller space in order to perform the quantization in the smaller dimension data space. At the cost of an extra gradient computation, the gradient estimate can be refined by conveying the difference between the gradient at the quantized data point and the original gradient using a small number of bits. Lastly, in order to save communication, our approach adds a layer that decides whether to transmit a quantized data sample or not based on its importance for learning. We analyze the convergence of the proposed approach for smooth convex and non-convex objective functions and show that we can achieve order optimal convergence rates with communication that mostly depends on the data rather than the model (gradient) dimension. We use our proposed algorithm to train ResNet models on the CIFAR-10 and ImageNet datasets, and show that we can achieve an order of magnitude savings over gradient compression methods. These communication savings come at the cost of increasing computation at the learning agent, and thus our approach is beneficial in scenarios where communication load is the main problem.},
  keywords={},
  doi={10.1109/JSAIT.2021.3105359},
  ISSN={2641-8770},
  month={Sep.},
  type={2},
  tags={journal,DML,CEDL},
  url_arxiv={https://arxiv.org/abs/2012.07913},
}

We consider machine learning applications that train a model by leveraging data distributed over a trusted network, where communication constraints can create a performance bottleneck. A number of recent approaches propose to overcome this bottleneck through compression of gradient updates. However, as models become larger, so does the size of the gradient updates. In this paper, we propose an alternate approach to learn from distributed data that quantizes data instead of gradients, and can support learning over applications where the size of gradient updates is prohibitive. Our approach leverages the dependency of the computed gradient on data samples, which lie in a much smaller space in order to perform the quantization in the smaller dimension data space. At the cost of an extra gradient computation, the gradient estimate can be refined by conveying the difference between the gradient at the quantized data point and the original gradient using a small number of bits. Lastly, in order to save communication, our approach adds a layer that decides whether to transmit a quantized data sample or not based on its importance for learning. We analyze the convergence of the proposed approach for smooth convex and non-convex objective functions and show that we can achieve order optimal convergence rates with communication that mostly depends on the data rather than the model (gradient) dimension. We use our proposed algorithm to train ResNet models on the CIFAR-10 and ImageNet datasets, and show that we can achieve an order of magnitude savings over gradient compression methods. These communication savings come at the cost of increasing computation at the learning agent, and thus our approach is beneficial in scenarios where communication load is the main problem.

@ARTICLE{9609622,
  author={Nikolopoulos, Pavlos and Srinivasavaradhan, Sundara Rajan and Fragouli, Christina and Diggavi, Suhas N.},
  journal={IEEE BITS the Information Theory Magazine}, 
  title={Group Testing for Community Infections}, 
  year={2021},
  volume={1},
  number={1},
  pages={57-68},
  abstract={Group testing is the technique of pooling together diagnostic samples in order to increase the efficiency of medical testing. Traditionally, works in group testing assume that the infections are i.i.d. However, contagious diseases like COVID-19 are governed by community spread and hence the infections are correlated. This survey presents an overview of recent research progress that leverages the community structure to further improve the efficiency of group testing. We show that taking into account the side-information provided by the community structure may lead to significant savings—up to 60\% fewer tests compared to traditional test designs. We review lower bounds and new approaches to encoding and decoding algorithms that take into account the community structure and integrate group testing into epidemiological modeling. Finally, we also discuss a few important open questions in this space.},
  keywords={},
  doi={10.1109/MBITS.2021.3126244},
  ISSN={2692-4110},
  month={Sep.},
  tags={journal,PET},
  type={2},
  }

Group testing is the technique of pooling together diagnostic samples in order to increase the efficiency of medical testing. Traditionally, works in group testing assume that the infections are i.i.d. However, contagious diseases like COVID-19 are governed by community spread and hence the infections are correlated. This survey presents an overview of recent research progress that leverages the community structure to further improve the efficiency of group testing. We show that taking into account the side-information provided by the community structure may lead to significant savings—up to 60% fewer tests compared to traditional test designs. We review lower bounds and new approaches to encoding and decoding algorithms that take into account the community structure and integrate group testing into epidemiological modeling. Finally, we also discuss a few important open questions in this space.

@article{10.1093/bioinformatics/btaa875,
 author = {Joshi, Dhaivat and Mao, Shunfu and Kannan, Sreeram and Diggavi, Suhas},
 title = "{QAlign: aligning nanopore reads accurately using current-level modeling}",
 journal = {Bioinformatics},
 volume = {37},
 number = {5},
 pages = {625-633},
 year = {2020},
 month = {12},
 abstract = "{Efficient and accurate alignment of DNA/RNA sequence reads to each other or to a reference genome/transcriptome is an important problem in genomic analysis. Nanopore sequencing has emerged as a major sequencing technology and many long-read aligners have been designed for aligning nanopore reads. However, the high error rate makes accurate and efficient alignment difficult. Utilizing the noise and error characteristics inherent in the sequencing process properly can play a vital role in constructing a robust aligner. In this article, we design QAlign, a pre-processor that can be used with any long-read aligner for aligning long reads to a genome/transcriptome or to other long reads. The key idea in QAlign is to convert the nucleotide reads into discretized current levels that capture the error modes of the nanopore sequencer before running it through a sequence aligner.We show that QAlign is able to improve alignment rates from around 80\\% up to 90\\% with nanopore reads when aligning to the genome. We also show that QAlign improves the average overlap quality by 9.2, 2.5 and 10.8\\% in three real datasets for read-to-read alignment. Read-to-transcriptome alignment rates are improved from 51.6\\% to 75.4\\% and 82.6\\% to 90\\% in two real datasets.https://github.com/joshidhaivat/QAlign.git.Supplementary data are available at Bioinformatics online.}",
 issn = {1367-4803},
 doi = {10.1093/bioinformatics/btaa875},
 tags = {journal,BioInf,NDS},
 title = {{QAlign: Aligning nanopore reads accurately using current-level modeling}},
 type = {2},
 url_biorxiv = {https://www.biorxiv.org/content/10.1101/862813v2},
 year = {2021}
}

Efficient and accurate alignment of DNA/RNA sequence reads to each other or to a reference genome/transcriptome is an important problem in genomic analysis. Nanopore sequencing has emerged as a major sequencing technology and many long-read aligners have been designed for aligning nanopore reads. However, the high error rate makes accurate and efficient alignment difficult. Utilizing the noise and error characteristics inherent in the sequencing process properly can play a vital role in constructing a robust aligner. In this article, we design QAlign, a pre-processor that can be used with any long-read aligner for aligning long reads to a genome/transcriptome or to other long reads. The key idea in QAlign is to convert the nucleotide reads into discretized current levels that capture the error modes of the nanopore sequencer before running it through a sequence aligner.We show that QAlign is able to improve alignment rates from around 80\% up to 90\% with nanopore reads when aligning to the genome. We also show that QAlign improves the average overlap quality by 9.2, 2.5 and 10.8\% in three real datasets for read-to-read alignment. Read-to-transcriptome alignment rates are improved from 51.6\% to 75.4\% and 82.6\% to 90\% in two real datasets.https://github.com/joshidhaivat/QAlign.git.Supplementary data are available at Bioinformatics online.

@article{agarwal2020distortion,
 abstract = {In Cyber-Physical Systems (CPS), inference based on communicated data is of critical significance as it can be used to manipulate or damage the control operations by adversaries. This calls for efficient mechanisms for secure transmission of data since control systems are becoming increasingly distributed over larger geographical areas. Distortion based security, recently proposed as one candidate for secure transmissions in CPS, is not only more appropriate for these applications but also quite frugal in terms of prior requirements on shared keys. In this paper, we propose distortion-based metrics to protect CPS communication and show that it is possible to confuse adversaries with just a few bits of pre-shared keys. In particular, we will show that a linear dynamical system can communicate its state in a manner that prevents an eavesdropper from accurately learning the state.},
 author = {Agarwal, G. K. and Karmoose, M. and Diggavi, S. and Fragouli, C. and Tabuada, P.},
 doi = {10.1109/TAC.2020.3006814},
 journal = {IEEE Transactions on Automatic Control},
 tags = {journal,CPS},
 title = {Distortion based Light-weight Security for Cyber-Physical Systems},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/2006.15998},
 year = {2021},
 volume={66},
 number={4},
 pages={1588-1601},
}

In Cyber-Physical Systems (CPS), inference based on communicated data is of critical significance as it can be used to manipulate or damage the control operations by adversaries. This calls for efficient mechanisms for secure transmission of data since control systems are becoming increasingly distributed over larger geographical areas. Distortion based security, recently proposed as one candidate for secure transmissions in CPS, is not only more appropriate for these applications but also quite frugal in terms of prior requirements on shared keys. In this paper, we propose distortion-based metrics to protect CPS communication and show that it is possible to confuse adversaries with just a few bits of pre-shared keys. In particular, we will show that a linear dynamical system can communicate its state in a manner that prevents an eavesdropper from accurately learning the state.

@article{DBLP:journals/corr/abs-1907-02664,
 abstract = {We study distributed optimization in the presence of Byzantine adversaries, where both data and computation are distributed among m worker machines, t of which may be corrupt. The compromised nodes may collaboratively and arbitrarily deviate from their pre-specified programs, and a designated (master) node iteratively computes the model/parameter vector for generalized linear models. In this work, we primarily focus on two iterative algorithms: Proximal Gradient Descent (PGD) and Coordinate Descent (CD). Gradient descent (GD) is a special case of these algorithms. PGD is typically used in the data-parallel setting, where data is partitioned across different samples, whereas, CD is used in the model-parallelism setting, where data is partitioned across the parameter space.  In this paper, we propose a method based on data encoding and error correction over real numbers to combat adversarial attacks. We can tolerate up to t≤⌊m−12⌋ corrupt worker nodes, which is information-theoretically optimal. We give deterministic guarantees, and our method does not assume any probability distribution on the data. We develop a {\em sparse} encoding scheme which enables computationally efficient data encoding and decoding. We demonstrate a trade-off between the corruption threshold and the resource requirements (storage, computational, and communication complexity). As an example, for t≤m3, our scheme incurs only a {\em constant} overhead on these resources, over that required by the plain distributed PGD/CD algorithms which provide no adversarial protection. To the best of our knowledge, ours is the first paper that makes CD secure against adversarial attacks.  Our encoding scheme extends efficiently to the data streaming model and for stochastic gradient descent (SGD). We also give experimental results to show the efficacy of our proposed schemes.},
 author = {Deepesh Data and
Linqi Song and
Suhas N. Diggavi},
 eprint = {1907.02664},
 journal = {IEEE Transactions on Information Theory},
 tags = {journal,SDL,DML},
 title = {Data Encoding for Byzantine-Resilient Distributed Optimization},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1907.02664},
 volume = {},
 year = {2021},
 volume={67},
 number={2},
 pages={1117-1140},
 doi = {10.1109/TIT.2020.3035868},
 ISSN={1557-9654},
 month={Feb},
}

We study distributed optimization in the presence of Byzantine adversaries, where both data and computation are distributed among m worker machines, t of which may be corrupt. The compromised nodes may collaboratively and arbitrarily deviate from their pre-specified programs, and a designated (master) node iteratively computes the model/parameter vector for generalized linear models. In this work, we primarily focus on two iterative algorithms: Proximal Gradient Descent (PGD) and Coordinate Descent (CD). Gradient descent (GD) is a special case of these algorithms. PGD is typically used in the data-parallel setting, where data is partitioned across different samples, whereas, CD is used in the model-parallelism setting, where data is partitioned across the parameter space. In this paper, we propose a method based on data encoding and error correction over real numbers to combat adversarial attacks. We can tolerate up to t≤⌊m−12⌋ corrupt worker nodes, which is information-theoretically optimal. We give deterministic guarantees, and our method does not assume any probability distribution on the data. We develop a \em sparse encoding scheme which enables computationally efficient data encoding and decoding. We demonstrate a trade-off between the corruption threshold and the resource requirements (storage, computational, and communication complexity). As an example, for t≤m3, our scheme incurs only a \em constant overhead on these resources, over that required by the plain distributed PGD/CD algorithms which provide no adversarial protection. To the best of our knowledge, ours is the first paper that makes CD secure against adversarial attacks. Our encoding scheme extends efficiently to the data streaming model and for stochastic gradient descent (SGD). We also give experimental results to show the efficacy of our proposed schemes.

@article{srinivasavaradhan2020algorithms,
 abstract = {Recent advances in DNA sequencing technology and DNA storage systems have rekindled the interest in deletion channels. Multiple recent works have looked at variants of sequence reconstruction over a single and over multiple deletion channels, a notoriously difficult problem due to its highly combinatorial nature. Although works in theoretical computer science have provided algorithms which guarantee perfect reconstruction with multiple independent observations from the deletion channel, they are only applicable in the large blocklength regime and more restrictively, when the number of observations is also large. Indeed, with only a few observations, perfect reconstruction of the input sequence may not even be possible in most cases. In such situations, maximum likelihood (ML) and maximum aposteriori (MAP) estimates for the deletion channels are natural questions that arise and these have remained open to the best of our knowledge. In this work, we take steps to answer the two aforementioned questions. Specifically: 1. We show that solving for the ML estimate over the single deletion channel (which can be cast as a discrete optimization problem) is equivalent to solving its relaxation, a continuous optimization problem; 2. We exactly compute the symbolwise posterior distributions (under some assumptions on the priors) for both the single as well as multiple deletion channels. As part of our contributions, we also introduce tools to visualize and analyze error events, which we believe could be useful in other related problems concerning deletion channels.},
 author = {Srinivasavaradhan, Sundara Rajan and Du, Michelle and Diggavi, Suhas and Fragouli, Christina},
 journal = {IEEE Transactions on Information Theory},
 volume={67},
 number={6},
 pages={3389-3410},
 doi={10.1109/TIT.2020.3033513},
 tags = {journal,BioInf,IT,NDS},
 title = {Algorithms for reconstruction over single and multiple deletion channels},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/2005.14388},
 ISSN={1557-9654},
 month={June},
 year = {2021}
}

Recent advances in DNA sequencing technology and DNA storage systems have rekindled the interest in deletion channels. Multiple recent works have looked at variants of sequence reconstruction over a single and over multiple deletion channels, a notoriously difficult problem due to its highly combinatorial nature. Although works in theoretical computer science have provided algorithms which guarantee perfect reconstruction with multiple independent observations from the deletion channel, they are only applicable in the large blocklength regime and more restrictively, when the number of observations is also large. Indeed, with only a few observations, perfect reconstruction of the input sequence may not even be possible in most cases. In such situations, maximum likelihood (ML) and maximum aposteriori (MAP) estimates for the deletion channels are natural questions that arise and these have remained open to the best of our knowledge. In this work, we take steps to answer the two aforementioned questions. Specifically: 1. We show that solving for the ML estimate over the single deletion channel (which can be cast as a discrete optimization problem) is equivalent to solving its relaxation, a continuous optimization problem; 2. We exactly compute the symbolwise posterior distributions (under some assumptions on the priors) for both the single as well as multiple deletion channels. As part of our contributions, we also introduce tools to visualize and analyze error events, which we believe could be useful in other related problems concerning deletion channels.

@INPROCEEDINGS{9500791,
  author={Nikolopoulos, Pavlos and Srinivasavaradhan, Sundara Rajan and Guo, Tao and Fragouli, Christina and Diggavi, Suhas},
  booktitle={ICC 2021 - IEEE International Conference on Communications}, 
  title={Group testing for overlapping communities}, 
  year={2021},
  volume={},
  number={},
  pages={1-7},
  abstract={In this paper, we propose algorithms that leverage a known community structure to make group testing more efficient. We consider a population organized in connected communities: each individual participates in one or more communities, and the infection probability of each individual depends on the communities (s)he participates in. Use cases include students who participate in several classes, and workers who share common spaces. Group testing reduces the number of tests needed to identify the infected individuals by pooling diagnostic samples and testing them together. We show that making testing algorithms aware of the community structure, can significantly reduce the number of tests needed both for adaptive and non-adaptive group testing.},
  keywords={},
  doi={10.1109/ICC42927.2021.9500791},
  ISSN={1938-1883},
  month={June},
 tags = {conf,PET},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/2012.02804},
 year = {2021},
}

In this paper, we propose algorithms that leverage a known community structure to make group testing more efficient. We consider a population organized in connected communities: each individual participates in one or more communities, and the infection probability of each individual depends on the communities (s)he participates in. Use cases include students who participate in several classes, and workers who share common spaces. Group testing reduces the number of tests needed to identify the infected individuals by pooling diagnostic samples and testing them together. We show that making testing algorithms aware of the community structure, can significantly reduce the number of tests needed both for adaptive and non-adaptive group testing.

@INPROCEEDINGS{9517986,
  author={Singh, Navjot and Data, Deepesh and George, Jemin and Diggavi, Suhas},
  booktitle={2021 IEEE International Symposium on Information Theory (ISIT)}, 
  title={SQuARM-SGD: Communication-Efficient Momentum SGD for Decentralized Optimization}, 
  year={2021},
  volume={},
  number={},
  pages={1212-1217},
  abstract={In this paper, we propose and analyze SQuARM-SGD, a communication-efficient algorithm for decentralized training of large-scale machine learning models over a network. In SQuARM-SGD, each node performs a fixed number of local SGD steps using Nesterov's momentum and then sends sparsified and quantized updates to its neighbors regulated by a locally computable triggering criterion. We provide convergence guarantees of our algorithm for general smooth objectives, which, to the best of our knowledge, is the first theoretical analysis for compressed decentralized SGD with momentum updates. We show that SQuARM-SGD converges at rate, matching that of vanilla distributed SGD. We empirically show that SQuARM-SGD saves significantly in total communicated bits over state-of-the-art without sacrificing much on accuracy.},
  keywords={},
  doi={10.1109/ISIT45174.2021.9517986},
  ISSN={},
  month={July},
  tags = {conf,CEDL,DML},
  type = {4},
  url_arxiv = {https://arxiv.org/abs/2005.07041},
  }

In this paper, we propose and analyze SQuARM-SGD, a communication-efficient algorithm for decentralized training of large-scale machine learning models over a network. In SQuARM-SGD, each node performs a fixed number of local SGD steps using Nesterov's momentum and then sends sparsified and quantized updates to its neighbors regulated by a locally computable triggering criterion. We provide convergence guarantees of our algorithm for general smooth objectives, which, to the best of our knowledge, is the first theoretical analysis for compressed decentralized SGD with momentum updates. We show that SQuARM-SGD converges at rate, matching that of vanilla distributed SGD. We empirically show that SQuARM-SGD saves significantly in total communicated bits over state-of-the-art without sacrificing much on accuracy.

@article{girgis2021renyidiff,
  title={Renyi Differential Privacy of the Subsampled Shuffle Model in Distributed Learning},
  author={Girgis, Antonious M and Data, Deepesh and Diggavi, Suhas},
  journal={Advances in Neural Information Processing Systems},
  volume={34},
  year={2021},
  tags={conf,PDL,DML},
  type={4},
  url_arxiv={https://arxiv.org/abs/2107.08763},
  abstract={We study privacy in a distributed learning framework, where clients collaboratively build a learning model iteratively through interactions with a server from whom we need privacy. Motivated by stochastic optimization and the federated learning (FL) paradigm, we focus on the case where a small fraction of data samples are randomly sub-sampled in each round to participate in the learning process, which also enables privacy amplification. To obtain even stronger local privacy guarantees, we study this in the shuffle privacy model, where each client randomizes its response using a local differentially private (LDP) mechanism and the server only receives a random permutation (shuffle) of the clients’ responses without their association to each client. The principal result of this paper is a privacy-optimization performance trade-off for discrete randomization mechanisms in this sub-sampled shuffle privacy model. This is enabled through a new theoretical technique to analyze the Renyi Differential Privacy (RDP) of the sub-sampled shuffle model. We numerically demonstrate that, for important regimes, with composition our bound yields significant improvement in privacy guarantee over the state-of-the-art approximate Differential Privacy (DP) guarantee (with strong composition) for sub-sampled shuffled models. We also demonstrate numerically significant improvement in privacy-learning performance operating point using real data sets.},
}

We study privacy in a distributed learning framework, where clients collaboratively build a learning model iteratively through interactions with a server from whom we need privacy. Motivated by stochastic optimization and the federated learning (FL) paradigm, we focus on the case where a small fraction of data samples are randomly sub-sampled in each round to participate in the learning process, which also enables privacy amplification. To obtain even stronger local privacy guarantees, we study this in the shuffle privacy model, where each client randomizes its response using a local differentially private (LDP) mechanism and the server only receives a random permutation (shuffle) of the clients’ responses without their association to each client. The principal result of this paper is a privacy-optimization performance trade-off for discrete randomization mechanisms in this sub-sampled shuffle privacy model. This is enabled through a new theoretical technique to analyze the Renyi Differential Privacy (RDP) of the sub-sampled shuffle model. We numerically demonstrate that, for important regimes, with composition our bound yields significant improvement in privacy guarantee over the state-of-the-art approximate Differential Privacy (DP) guarantee (with strong composition) for sub-sampled shuffled models. We also demonstrate numerically significant improvement in privacy-learning performance operating point using real data sets.

@inproceedings{10.1145/3460120.3484794,
author = {Girgis, Antonious M. and Data, Deepesh and Diggavi, Suhas and Suresh, Ananda Theertha and Kairouz, Peter},
title = {On the Renyi Differential Privacy of the Shuffle Model},
year = {2021},
isbn = {9781450384544},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3460120.3484794},
doi = {10.1145/3460120.3484794},
abstract = {The central question studied in this paper is R\'{e}nyi Differential Privacy (RDP) guarantees for general discrete local randomizers in the shuffle privacy model. In the shuffle model, each of the n clients randomizes its response using a local differentially private (LDP) mechanism and the untrusted server only receives a random permutation (shuffle) of the client responses without association to each client. The principal result in this paper is the first direct RDP bounds for general discrete local randomization in the shuffle privacy model, and we develop new analysis techniques for deriving our results which could be of independent interest. In applications, such an RDP guarantee is most useful when we use it for composing several private interactions. We numerically demonstrate that, for important regimes, with composition our bound yields an improvement in privacy guarantee by a factor of $8times$ over the state-of-the-art approximate Differential Privacy (DP) guarantee (with standard composition) for shuffle models. Moreover, combining with Poisson subsampling, our result leads to at least $10times$ improvement over subsampled approximate DP with standard composition.},
booktitle = {Proceedings of the 2021 ACM SIGSAC Conference on Computer and Communications Security},
pages = {2321–2341},
numpages = {21},
keywords = {distributed learning, renyi divergence, privacy amplification via shuffling, privacy composition, differential privacy},
location = {Virtual Event, Republic of Korea},
series = {CCS '21},
  url_arxiv = {https://arxiv.org/abs/2105.05180},
  tags = {conf,PDL,DML},
  type = {4},
}

The central question studied in this paper is Rényi Differential Privacy (RDP) guarantees for general discrete local randomizers in the shuffle privacy model. In the shuffle model, each of the n clients randomizes its response using a local differentially private (LDP) mechanism and the untrusted server only receives a random permutation (shuffle) of the client responses without association to each client. The principal result in this paper is the first direct RDP bounds for general discrete local randomization in the shuffle privacy model, and we develop new analysis techniques for deriving our results which could be of independent interest. In applications, such an RDP guarantee is most useful when we use it for composing several private interactions. We numerically demonstrate that, for important regimes, with composition our bound yields an improvement in privacy guarantee by a factor of $8times$ over the state-of-the-art approximate Differential Privacy (DP) guarantee (with standard composition) for shuffle models. Moreover, combining with Poisson subsampling, our result leads to at least $10times$ improvement over subsampled approximate DP with standard composition.

@article{ozkara2021quped,
  title={QuPeD: Quantized Personalization via Distillation with Applications to Federated Learning},
  author={Ozkara, Kaan and Singh, Navjot and Data, Deepesh and Diggavi, Suhas},
  journal={Advances in Neural Information Processing Systems},
  volume={34},
  year={2021},
  tags = {conf,CEDL,DML,PFL},
  url_arxiv = {https://arxiv.org/abs/2107.13892},
  type = {4},
  abstract={Traditionally, federated learning (FL) aims to train a single global model while collaboratively using multiple clients and a server. Two natural challenges that FL algorithms face are heterogeneity in data across clients and collaboration of clients with diverse resources. In this work, we introduce a quantized and personalized FL algorithm QuPeD that facilitates collective (personalized model compression) training via knowledge distillation (KD) among clients who have access to heterogeneous data and resources. For personalization, we allow clients to learn compressed personalized models with different quantization parameters and model dimensions/structures. Towards this, first we propose an algorithm for learning quantized models through a relaxed optimization problem, where quantization values are also optimized over. When each client participating in the (federated) learning process has different requirements for the compressed model (both in model dimension and precision), we formulate a compressed personalization framework by introducing knowledge distillation loss for local client objectives collaborating through a global model. We develop an alternating proximal gradient update for solving this compressed personalization problem, and analyze its convergence properties. Numerically, we validate that QuPeD outperforms competing personalized FL methods, FedAvg, and local training of clients in various heterogeneous settings.},
}

Traditionally, federated learning (FL) aims to train a single global model while collaboratively using multiple clients and a server. Two natural challenges that FL algorithms face are heterogeneity in data across clients and collaboration of clients with diverse resources. In this work, we introduce a quantized and personalized FL algorithm QuPeD that facilitates collective (personalized model compression) training via knowledge distillation (KD) among clients who have access to heterogeneous data and resources. For personalization, we allow clients to learn compressed personalized models with different quantization parameters and model dimensions/structures. Towards this, first we propose an algorithm for learning quantized models through a relaxed optimization problem, where quantization values are also optimized over. When each client participating in the (federated) learning process has different requirements for the compressed model (both in model dimension and precision), we formulate a compressed personalization framework by introducing knowledge distillation loss for local client objectives collaborating through a global model. We develop an alternating proximal gradient update for solving this compressed personalization problem, and analyze its convergence properties. Numerically, we validate that QuPeD outperforms competing personalized FL methods, FedAvg, and local training of clients in various heterogeneous settings.

@article{data2020byzantine2,
 abstract = {We study stochastic gradient descent (SGD) with local iterations in the presence of malicious/Byzantine clients, motivated by the federated learning. The clients, instead of communicating with the central server in every iteration, maintain their local models, which they update by taking several SGD iterations based on their own datasets and then communicate the net update with the server, thereby achieving communication-efficiency. Furthermore, only a subset of clients communicate with the server, and this subset may be different at different synchronization times. The Byzantine clients may collaborate and send arbitrary vectors to the server to disrupt the learning process. To combat the adversary, we employ an efficient high-dimensional robust mean estimation algorithm from Steinhardt et al.~\cite[ITCS 2018]{Resilience_SCV18} at the server to filter-out corrupt vectors; and to analyze the outlier-filtering procedure, we develop a novel matrix concentration result that may be of independent interest. 
We provide convergence analyses for strongly-convex and non-convex smooth objectives in the heterogeneous data setting, where different clients may have different local datasets, and we do not make any probabilistic assumptions on data generation. We believe that ours is the first Byzantine-resilient algorithm and analysis with local iterations. We derive our convergence results under minimal assumptions of bounded variance for SGD and bounded gradient dissimilarity (which captures heterogeneity among local datasets). We also extend our results to the case when clients compute full-batch gradients.},
 author = {Data, Deepesh and Diggavi, Suhas},
 journal = {Proceedings International Conference on Machine Learning (ICML)},
 tags = {conf,SDL,DML},
 title = {Byzantine-Resilient High-Dimensional SGD with Local Iterations on Heterogeneous Data},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/2006.13041},
 year = {2021}
}

We study stochastic gradient descent (SGD) with local iterations in the presence of malicious/Byzantine clients, motivated by the federated learning. The clients, instead of communicating with the central server in every iteration, maintain their local models, which they update by taking several SGD iterations based on their own datasets and then communicate the net update with the server, thereby achieving communication-efficiency. Furthermore, only a subset of clients communicate with the server, and this subset may be different at different synchronization times. The Byzantine clients may collaborate and send arbitrary vectors to the server to disrupt the learning process. To combat the adversary, we employ an efficient high-dimensional robust mean estimation algorithm from Steinhardt et al. i̧te[ITCS 2018]Resilience_SCV18 at the server to filter-out corrupt vectors; and to analyze the outlier-filtering procedure, we develop a novel matrix concentration result that may be of independent interest. We provide convergence analyses for strongly-convex and non-convex smooth objectives in the heterogeneous data setting, where different clients may have different local datasets, and we do not make any probabilistic assumptions on data generation. We believe that ours is the first Byzantine-resilient algorithm and analysis with local iterations. We derive our convergence results under minimal assumptions of bounded variance for SGD and bounded gradient dissimilarity (which captures heterogeneity among local datasets). We also extend our results to the case when clients compute full-batch gradients.

@article{nikolopoulos2021community,
  title = {Group testing for connected communities},
  author = {Nikolopoulos, Pavlos and Guo, Tao and Srinivasavaradhan, Sundara Rajan and Fragouli, Christina and Diggavi, Suhas},
  journal = {Proceedings of The 24th International Conference on Artificial Intelligence and Statistics},
  year = {2021},
  pages =   {2341--2349},
  type = {4},
  tags = {conf,PET},
  url_arxiv = {https://arxiv.org/abs/2007.08111},
  volume =   {130},
  series =   {Proceedings of Machine Learning Research},
  month =   {13--15 Apr},
  publisher =    {PMLR},
  url = {http://proceedings.mlr.press/v130/nikolopoulos21a.html},
  abstract =   { In this paper, we propose algorithms that leverage a known community structure to make group testing more efficient. We consider a population organized in disjoint communities: each individual participates in a community, and its infection probability depends on the community (s)he participates in. Use cases include families, students who participate in several classes, and workers who share common spaces. Group testing reduces the number of tests needed to identify the infected individuals by pooling diagnostic samples and testing them together. We show that if we design the testing strategy taking into account the community structure, we can significantly reduce the number of tests needed for adaptive and non-adaptive group testing, and can improve the reliability in cases where tests are noisy.}
}

In this paper, we propose algorithms that leverage a known community structure to make group testing more efficient. We consider a population organized in disjoint communities: each individual participates in a community, and its infection probability depends on the community (s)he participates in. Use cases include families, students who participate in several classes, and workers who share common spaces. Group testing reduces the number of tests needed to identify the infected individuals by pooling diagnostic samples and testing them together. We show that if we design the testing strategy taking into account the community structure, we can significantly reduce the number of tests needed for adaptive and non-adaptive group testing, and can improve the reliability in cases where tests are noisy.

@article{girgis2020shuffledpreprint,
 abstract = {We consider a distributed empirical risk minimization (ERM) optimization problem with communication efficiency and privacy requirements, motivated by the federated learning (FL) framework. Unique challenges to the traditional ERM problem in the context of FL include (i) need to provide privacy guarantees on clients' data, (ii) compress the communication between clients and the server, since clients might have low-bandwidth links, (iii) work with a dynamic client population at each round of communication between the server and the clients, as a small fraction of clients are sampled at each round. To address these challenges we develop (optimal) communication-efficient schemes for private mean estimation for several ℓp spaces, enabling efficient gradient aggregation for each iteration of the optimization solution of the ERM. We also provide lower and upper bounds for mean estimation with privacy and communication constraints for arbitrary ℓp spaces. To get the overall communication, privacy, and optimization performance operation point, we combine this with privacy amplification opportunities inherent to this setup. Our solution takes advantage of the inherent privacy amplification provided by client sampling and data sampling at each client (through Stochastic Gradient Descent) as well as the recently developed privacy framework using anonymization, which effectively presents to the server responses that are randomly shuffled with respect to the clients. Putting these together, we demonstrate that one can get the same privacy, optimization-performance operating point developed in recent methods that use full-precision communication, but at a much lower communication cost, i.e., effectively getting communication efficiency for "free".},
 author = {Girgis, Antonious M and Data, Deepesh and Diggavi, Suhas and Kairouz, Peter and Suresh, Ananda Theertha},
 journal = {Proceedings of The 24th International Conference on Artificial Intelligence and Statistics},
 pages={2521--2529},
 tags = {conf,DML,PDL},
 organization={PMLR},
 title = {Shuffled Model of Federated Learning: Privacy, Communication and Accuracy Trade-offs},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/2008.07180},
 year = {2021},
}

We consider a distributed empirical risk minimization (ERM) optimization problem with communication efficiency and privacy requirements, motivated by the federated learning (FL) framework. Unique challenges to the traditional ERM problem in the context of FL include (i) need to provide privacy guarantees on clients' data, (ii) compress the communication between clients and the server, since clients might have low-bandwidth links, (iii) work with a dynamic client population at each round of communication between the server and the clients, as a small fraction of clients are sampled at each round. To address these challenges we develop (optimal) communication-efficient schemes for private mean estimation for several ℓp spaces, enabling efficient gradient aggregation for each iteration of the optimization solution of the ERM. We also provide lower and upper bounds for mean estimation with privacy and communication constraints for arbitrary ℓp spaces. To get the overall communication, privacy, and optimization performance operation point, we combine this with privacy amplification opportunities inherent to this setup. Our solution takes advantage of the inherent privacy amplification provided by client sampling and data sampling at each client (through Stochastic Gradient Descent) as well as the recently developed privacy framework using anonymization, which effectively presents to the server responses that are randomly shuffled with respect to the clients. Putting these together, we demonstrate that one can get the same privacy, optimization-performance operating point developed in recent methods that use full-precision communication, but at a much lower communication cost, i.e., effectively getting communication efficiency for "free".

@inproceedings{girgis2021differentially,
  title={Differentially Private Federated Learning with Shuffling and Client Self-Sampling},
  author={Girgis, Antonious and Data, Deepesh and Diggavi, Suhas},
  booktitle={IEEE International Symposium on Information Theory (ISIT)},
  year={2021},
  type={4},
  tags={conf,PDL,DML},
}

@inproceedings{sundar2021entropy,
  title={An Entropy Reduction Approach to Continual Testing},
  author={Srinivasavaradhan, Sundar and Nikolopoulos, Pavlos and Fragouli, Christina and Diggavi, Suhas N.},
  booktitle={IEEE International Symposium on Information Theory (ISIT)},
  year={2021},
  type={4},
  tags={conf,PET},
}

@ARTICLE{9527149,
  author={Diggavi, Suhas and Gamal, Abbas El and Kontoyiannis, Ioannis and Paranjpe, Ajit and Subbarao, Murali},
  journal={IEEE BITS the Information Theory Magazine}, 
  title={Joy Thomas: Legacy, Foundation and the IT Society}, 
  year={2021},
  volume={1},
  number={1},
  pages={85-88},
  abstract={In this article, we commemorate Joy Thomas, who sadly passed away in 2020. Joy is best known for his seminal textbook Elements of Information Theory, coauthored with his Ph.D. advisor, the late Thomas Cover. We also catalog some of Joy’s technical contributions to information theory and to its practice. Joy also made an impact on industry through several successful startups. In recognition of Joy’s contribution to the exposition of information theory, the IEEE Information Theory Society Board of Governors has agreed to establish the Joy Thomas IEEE Information Theory Society Tutorial Paper Award. The Joy Thomas Foundation, established to honor Joy’s legacy of excellence and giving back, will sponsor this award.},
  keywords={},
  doi={10.1109/MBITS.2021.3108946},
  ISSN={2692-4110},
  month={Sep.},
  tags={perspectives},
  type={7},
  }

In this article, we commemorate Joy Thomas, who sadly passed away in 2020. Joy is best known for his seminal textbook Elements of Information Theory, coauthored with his Ph.D. advisor, the late Thomas Cover. We also catalog some of Joy’s technical contributions to information theory and to its practice. Joy also made an impact on industry through several successful startups. In recognition of Joy’s contribution to the exposition of information theory, the IEEE Information Theory Society Board of Governors has agreed to establish the Joy Thomas IEEE Information Theory Society Tutorial Paper Award. The Joy Thomas Foundation, established to honor Joy’s legacy of excellence and giving back, will sponsor this award.

@article{data2020byzantine1,
 abstract = {We study distributed stochastic gradient descent (SGD) in the master-worker architecture under Byzantine attacks. We consider the heterogeneous data model, where different workers may have different local datasets, and we do not make any probabilistic assumptions on data generation. At the core of our algorithm, we use the polynomial-time outlier-filtering procedure for robust mean estimation proposed by Steinhardt et al. (ITCS 2018) to filter-out corrupt gradients. In order to be able to apply their filtering procedure in our {\em heterogeneous} data setting where workers compute {\em stochastic} gradients, we derive a new matrix concentration result, which may be of independent interest. 
We provide convergence analyses for smooth strongly-convex and non-convex objectives. We derive our results under the bounded variance assumption on local stochastic gradients and a {\em deterministic} condition on datasets, namely, gradient dissimilarity; and for both these quantities, we provide concrete bounds in the statistical heterogeneous data model. We give a trade-off between the mini-batch size for stochastic gradients and the approximation error. Our algorithm can tolerate up to 14 fraction Byzantine workers. It can find approximate optimal parameters in the strongly-convex setting exponentially fast and reach to an approximate stationary point in the non-convex setting with a linear speed, thus, matching the convergence rates of vanilla SGD in the Byzantine-free setting. 
We also propose and analyze a Byzantine-resilient SGD algorithm with gradient compression, where workers send k random coordinates of their gradients. Under mild conditions, we show a d/k-factor saving in communication bits as well as decoding complexity over our compression-free algorithm without affecting its convergence rate (order-wise) and the approximation error.},
 author = {Data, Deepesh and Diggavi, Suhas},
 journal = {arXiv preprint arXiv:2005.07866},
 tags = {journalSub,SDL,DML},
 title = {Byzantine-Resilient SGD in High Dimensions on Heterogeneous Data},
 type = {1},
 url_arxiv = {https://arxiv.org/abs/2005.07866},
 year = {2020}
}

We study distributed stochastic gradient descent (SGD) in the master-worker architecture under Byzantine attacks. We consider the heterogeneous data model, where different workers may have different local datasets, and we do not make any probabilistic assumptions on data generation. At the core of our algorithm, we use the polynomial-time outlier-filtering procedure for robust mean estimation proposed by Steinhardt et al. (ITCS 2018) to filter-out corrupt gradients. In order to be able to apply their filtering procedure in our \em heterogeneous data setting where workers compute \em stochastic gradients, we derive a new matrix concentration result, which may be of independent interest. We provide convergence analyses for smooth strongly-convex and non-convex objectives. We derive our results under the bounded variance assumption on local stochastic gradients and a \em deterministic condition on datasets, namely, gradient dissimilarity; and for both these quantities, we provide concrete bounds in the statistical heterogeneous data model. We give a trade-off between the mini-batch size for stochastic gradients and the approximation error. Our algorithm can tolerate up to 14 fraction Byzantine workers. It can find approximate optimal parameters in the strongly-convex setting exponentially fast and reach to an approximate stationary point in the non-convex setting with a linear speed, thus, matching the convergence rates of vanilla SGD in the Byzantine-free setting. We also propose and analyze a Byzantine-resilient SGD algorithm with gradient compression, where workers send k random coordinates of their gradients. Under mild conditions, we show a d/k-factor saving in communication bits as well as decoding complexity over our compression-free algorithm without affecting its convergence rate (order-wise) and the approximation error.

@article{9023963,
 abstract = {We study the generalized degrees of freedom (gDoF) of block-fading noncoherent multiple input multiple output (MIMO) channels with asymmetric distributions of link strengths and a coherence time of T symbol durations. We derive the optimal signaling structure for communication for the asymmetric MIMO channel, which is distinct from that for the MIMO channel with independent and identically distributed (i.i.d.) links. We extend the existing results for the single input multiple output (SIMO) channel with i.i.d. links to the asymmetric case, proving that selecting the statistically best antenna is gDoFoptimal. Using the gDoF result for the SIMO channel, we prove that for T = 1, the gDoF is zero for MIMO channels with arbitrary link strengths. We show that selecting the statistically best antenna is gDoF-optimal for the multiple input single output (MISO) channel. We also derive the gDoF for the 2 x 2 MIMO channel with different exponents in the direct and cross links. In this setting, we show that it is always necessary to use both the antennas to achieve the gDoF, in contrast to the results for the 2 x 2 MIMO channel with i.i.d. links. We show that having weaker crosslinks, gives gDoF gain compared to the case with i.i.d. links. For the noncoherent MIMO channel with i.i.d. links, the traditional method of training each transmit antenna independently is degrees of freedom (DoF) optimal, whereas we observe that for the asymmetric 2 x 2 MIMO channel, the traditional training is not gDoF-optimal. We extend this observation to a larger MxM MIMO channel by demonstrating a strategy that can achieve larger gDoF than a traditional trainingbased method.},
 author = {J. {Sebastian} and S. N. {Diggavi}},
 doi = {10.1109/TIT.2020.2978183},
 issn = {1557-9654},
 journal = {IEEE Transactions on Information Theory},
 keywords = {MIMO communication;MISO communication;SIMO communication;transmitting antennas;generalized degrees freedom;noncoherent MIMO channel;asymmetric link strengths;block-fading noncoherent multiple input multiple output channels;asymmetric distributions;optimal signaling structure;asymmetric MIMO channel;single input multiple output channel;asymmetric case;SIMO channel;arbitrary link strengths;multiple input single output channel;direct links;cross links;gDoF gain;degrees of freedom optimal;MxM MIMO channel;MIMO communication;Signal to noise ratio;Fading channels;Transmitting antennas;Coherence;MISO communication;Degrees of freedom (DoF);multiple antennas;time-varying channels;noncoherent communication},
 month = {July},
 number = {7},
 pages = {4431-4448},
 tags = {journal,IT,ANIT,WiNetnew,NCWN,WNIF},
 title = {Generalized Degrees Freedom of Noncoherent MIMO Channels With Asymmetric Link Strengths},
 type = {2},
 volume = {66},
 year = {2020}
}

We study the generalized degrees of freedom (gDoF) of block-fading noncoherent multiple input multiple output (MIMO) channels with asymmetric distributions of link strengths and a coherence time of T symbol durations. We derive the optimal signaling structure for communication for the asymmetric MIMO channel, which is distinct from that for the MIMO channel with independent and identically distributed (i.i.d.) links. We extend the existing results for the single input multiple output (SIMO) channel with i.i.d. links to the asymmetric case, proving that selecting the statistically best antenna is gDoFoptimal. Using the gDoF result for the SIMO channel, we prove that for T = 1, the gDoF is zero for MIMO channels with arbitrary link strengths. We show that selecting the statistically best antenna is gDoF-optimal for the multiple input single output (MISO) channel. We also derive the gDoF for the 2 x 2 MIMO channel with different exponents in the direct and cross links. In this setting, we show that it is always necessary to use both the antennas to achieve the gDoF, in contrast to the results for the 2 x 2 MIMO channel with i.i.d. links. We show that having weaker crosslinks, gives gDoF gain compared to the case with i.i.d. links. For the noncoherent MIMO channel with i.i.d. links, the traditional method of training each transmit antenna independently is degrees of freedom (DoF) optimal, whereas we observe that for the asymmetric 2 x 2 MIMO channel, the traditional training is not gDoF-optimal. We extend this observation to a larger MxM MIMO channel by demonstrating a strategy that can achieve larger gDoF than a traditional trainingbased method.

@article{9046857,
 abstract = {We study the generalized degrees of freedom (gDoF) of the noncoherent diamond (parallel relay) wireless network with asymmetric distributions of link strengths. We use the noncoherent block-fading model introduced by Marzetta and Hochwald, where no channel state information is available at the transmitters or at the receivers and the channels remain constant for a coherence time of T symbol durations. We first derive an upper bound for the capacity of this channel and then derive the optimal structure for the solution of the upper bound optimization problem. Using the optimal structure, we solve the upper bound optimization problem in terms of its gDoF. Using insights from our upper bound signaling solution, we devise an achievability strategy based on a novel scheme that we call train-scale quantize-map-forward (TS-QMF). This scheme uses training in the links from the source to the relays, scaling and quantizing at the relays combined with nontraining-based schemes. We show the optimality of this scheme by comparing it to the upper bound in terms of the gDoF. In noncoherent point-to-point multiple-input-multiple-output (MIMO) channels, where the fading realization is unknown to the transmitter and the receiver, an important tradeoff between communication and channel learning was revealed by Zheng and Tse, by demonstrating that not all the available antennas might be used, as it is suboptimal to learn all their channel parameters. Our results in this paper for the diamond network demonstrate that in certain regimes of relative channel strengths, the gDoF-optimal scheme uses a subnetwork, demonstrating a similar tradeoff between channel learning and communication. In some regimes, it is gDoF-optimal to do relay selection, i.e., use a part of the network. In the other regimes, even when it is essential to use the entire network, it is suboptimal to learn the channel states for all the links in the network, i.e., traditional training-based schemes are suboptimal in these regimes.},
 author = {J. {Sebastian} and S. {Diggavi}},
 doi = {10.1109/TIT.2020.2983169},
 issn = {1557-9654},
 journal = {IEEE Transactions on Information Theory},
 keywords = {channel capacity;fading channels;learning (artificial intelligence);MIMO communication;optimisation;quantisation (signal);radio receivers;radio transmitters;relay networks (telecommunication);telecommunication computing;noncoherent point-to-point MIMO channel;noncoherent point-to-point multiple-input-multiple-output channel;TS-QMF;bound signaling solution;bound optimization problem;channel capacity;Hochwald;Marzetta;block-fading model;asymmetric distribution;noncoherent diamond wireless network;generalized degrees of freedom;noncoherent diamond network;gDoF-optimal scheme;train-scale quantize-map-forward;channel state information;Signal to noise ratio;Relays;Diamond;MIMO communication;Upper bound;Fading channels;Wireless networks;Noncoherent communication;degrees of freedom (DoF);relay channels;diamond network;time-varying channels},
 month = {Aug},
 number = {8},
 pages = {5228-5260},
 tags = {journal,IT,ANIT,WiNetnew,NCWN,WNIF},
 title = {Generalized Degrees of Freedom of Noncoherent Diamond Networks},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1802.02667},
 volume = {66},
 year = {2020}
}

We study the generalized degrees of freedom (gDoF) of the noncoherent diamond (parallel relay) wireless network with asymmetric distributions of link strengths. We use the noncoherent block-fading model introduced by Marzetta and Hochwald, where no channel state information is available at the transmitters or at the receivers and the channels remain constant for a coherence time of T symbol durations. We first derive an upper bound for the capacity of this channel and then derive the optimal structure for the solution of the upper bound optimization problem. Using the optimal structure, we solve the upper bound optimization problem in terms of its gDoF. Using insights from our upper bound signaling solution, we devise an achievability strategy based on a novel scheme that we call train-scale quantize-map-forward (TS-QMF). This scheme uses training in the links from the source to the relays, scaling and quantizing at the relays combined with nontraining-based schemes. We show the optimality of this scheme by comparing it to the upper bound in terms of the gDoF. In noncoherent point-to-point multiple-input-multiple-output (MIMO) channels, where the fading realization is unknown to the transmitter and the receiver, an important tradeoff between communication and channel learning was revealed by Zheng and Tse, by demonstrating that not all the available antennas might be used, as it is suboptimal to learn all their channel parameters. Our results in this paper for the diamond network demonstrate that in certain regimes of relative channel strengths, the gDoF-optimal scheme uses a subnetwork, demonstrating a similar tradeoff between channel learning and communication. In some regimes, it is gDoF-optimal to do relay selection, i.e., use a part of the network. In the other regimes, even when it is essential to use the entire network, it is suboptimal to learn the channel states for all the links in the network, i.e., traditional training-based schemes are suboptimal in these regimes.

@article{9057579,
 abstract = {Communication bottleneck has been identified as a significant issue in distributed optimization of large-scale learning models. Recently, several approaches to mitigate this problem have been proposed, including different forms of gradient compression or computing local models and mixing them iteratively. In this paper, we propose Qsparse-local-SGD algorithm, which combines aggressive sparsification with quantization and local computation along with error compensation, by keeping track of the difference between the true and compressed gradients. We propose both synchronous and asynchronous implementations of Qsparse-local-SGD. We analyze convergence for Qsparse-local-SGD in the distributed setting for smooth non-convex and convex objective functions. We demonstrate that Qsparse-local-SGD converges at the same rate as vanilla distributed SGD for many important classes of sparsifiers and quantizers. We use Qsparse-local-SGD to train ResNet-50 on ImageNet and show that it results in significant savings over the state-of-the-art, in the number of bits transmitted to reach target accuracy.},
 author = {D. {Basu} and D. {Data} and C. {Karakus} and S. N. {Diggavi}},
 doi = {10.1109/JSAIT.2020.2985917},
 issn = {2641-8770},
 journal = {IEEE Journal on Selected Areas in Information Theory},
 keywords = {concave programming;convex programming;gradient methods;learning (artificial intelligence);optimisation;stochastic processes;stochastic gradient descent;convex objective functions;nonconvex objective functions;error compensation;sparsification;Qsparse-local-SGD;distributed setting;compressed gradients;true gradients;gradient compression;large-scale learning models;distributed optimization;quantization;vanilla distributed SGD;Quantization (signal);Convergence;Computational modeling;Stochastic processes;Training;Peer-to-peer computing;Optimization;Distributed optimization and learning;stochastic optimization;communication efficient training methods},
 month = {May},
 number = {1},
 pages = {217-226},
 tags = {journal,CEDL,DML},
 title = {Qsparse-Local-SGD: Distributed SGD With Quantization, Sparsification, and Local Computations},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1906.02367},
 volume = {1},
 year = {2020}
}

Communication bottleneck has been identified as a significant issue in distributed optimization of large-scale learning models. Recently, several approaches to mitigate this problem have been proposed, including different forms of gradient compression or computing local models and mixing them iteratively. In this paper, we propose Qsparse-local-SGD algorithm, which combines aggressive sparsification with quantization and local computation along with error compensation, by keeping track of the difference between the true and compressed gradients. We propose both synchronous and asynchronous implementations of Qsparse-local-SGD. We analyze convergence for Qsparse-local-SGD in the distributed setting for smooth non-convex and convex objective functions. We demonstrate that Qsparse-local-SGD converges at the same rate as vanilla distributed SGD for many important classes of sparsifiers and quantizers. We use Qsparse-local-SGD to train ResNet-50 on ImageNet and show that it results in significant savings over the state-of-the-art, in the number of bits transmitted to reach target accuracy.

@article{9061031,
 abstract = {We consider the problem of distributed feature quantization, where the goal is to enable a pretrained classifier at a central node to carry out its classification on features that are gathered from distributed nodes through communication constrained channels. We propose the design of distributed quantization schemes specifically tailored to the classification task: unlike quantization schemes that help the central node reconstruct the original signal as accurately as possible, our focus is not reconstruction accuracy, but instead correct classification. Our work does not make any a priori distributional assumptions on the data, but instead uses training data for the quantizer design. Our main contributions include: we prove NP-hardness of finding optimal quantizers in the general case; we design an optimal scheme for a special case; we propose quantization algorithms, that leverage discrete neural representations and training data, and can be designed in polynomial-time for any number of features, any number of classes, and arbitrary division of features across the distributed nodes. We find that tailoring the quantizers to the classification task can offer significant savings: as compared to alternatives, we can achieve more than a factor of two reduction in terms of the number of bits communicated, for the same classification accuracy.},
 author = {O. A. {Hanna} and Y. H. {Ezzeldin} and T. {Sadjadpour} and C. {Fragouli} and S. {Diggavi}},
 doi = {10.1109/JSAIT.2020.2986467},
 issn = {2641-8770},
 journal = {IEEE Journal on Selected Areas in Information Theory},
 keywords = {computational complexity;optimisation;quantisation (signal);signal classification;signal reconstruction;signal representation;distributed feature quantization;pretrained classifier;central node;distributed nodes;communication constrained channels;classification task;quantizer design;quantization algorithms;neural representations;training data;distributed quantization;NP-hardness;signal reconstruction;Quantization (signal);Image reconstruction;Training;Testing;Information theory;Task analysis;Training data;Distributed quantization;inference;communication-bounded inference;quantization with deep learning},
 month = {May},
 number = {1},
 pages = {237-249},
 tags = {journal,DML,CEDL},
 title = {On Distributed Quantization for Classification},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1911.00216},
 volume = {1},
 year = {2020}
}

We consider the problem of distributed feature quantization, where the goal is to enable a pretrained classifier at a central node to carry out its classification on features that are gathered from distributed nodes through communication constrained channels. We propose the design of distributed quantization schemes specifically tailored to the classification task: unlike quantization schemes that help the central node reconstruct the original signal as accurately as possible, our focus is not reconstruction accuracy, but instead correct classification. Our work does not make any a priori distributional assumptions on the data, but instead uses training data for the quantizer design. Our main contributions include: we prove NP-hardness of finding optimal quantizers in the general case; we design an optimal scheme for a special case; we propose quantization algorithms, that leverage discrete neural representations and training data, and can be designed in polynomial-time for any number of features, any number of classes, and arbitrary division of features across the distributed nodes. We find that tailoring the quantizers to the classification task can offer significant savings: as compared to alternatives, we can achieve more than a factor of two reduction in terms of the number of bits communicated, for the same classification accuracy.

@article{girgis2020successive,
 abstract = {This work examines a novel question: how much randomness is needed to achieve local differential privacy (LDP)? A motivating scenario is providing {\em multiple levels of privacy} to multiple analysts, either for distribution or for heavy-hitter estimation, using the \emph{same} (randomized) output. We call this setting \emph{successive refinement of privacy}, as it provides hierarchical access to the raw data with different privacy levels. For example, the same randomized output could enable one analyst to reconstruct the input, while another can only estimate the distribution subject to LDP requirements. This extends the classical Shannon (wiretap) security setting to local differential privacy. We provide (order-wise) tight characterizations of privacy-utility-randomness trade-offs in several cases for distribution estimation, including the standard LDP setting under a randomness constraint. We also provide a non-trivial privacy mechanism for multi-level privacy. Furthermore, we show that we cannot reuse random keys over time while preserving privacy of each user.},
 author = {Girgis, Antonious M and Data, Deepesh and Chaudhuri, Kamalika and Fragouli, Christina and Diggavi, Suhas},
 journal = {IEEE Journal on Selected Areas in Information Theory (JSAIT)},
 tags = {journal,DML,PDL},
 title = {Successive Refinement of Privacy},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/2005.11651},
 year = {2020},
 doi={10.1109/JSAIT.2020.3040403},
 volume={1},
 number={3},
 pages={745-759},
}

This work examines a novel question: how much randomness is needed to achieve local differential privacy (LDP)? A motivating scenario is providing \em multiple levels of privacy to multiple analysts, either for distribution or for heavy-hitter estimation, using the \emphsame (randomized) output. We call this setting \emphsuccessive refinement of privacy, as it provides hierarchical access to the raw data with different privacy levels. For example, the same randomized output could enable one analyst to reconstruct the input, while another can only estimate the distribution subject to LDP requirements. This extends the classical Shannon (wiretap) security setting to local differential privacy. We provide (order-wise) tight characterizations of privacy-utility-randomness trade-offs in several cases for distribution estimation, including the standard LDP setting under a randomness constraint. We also provide a non-trivial privacy mechanism for multi-level privacy. Furthermore, we show that we cannot reuse random keys over time while preserving privacy of each user.

@article{SHOWKATBAKHSH2020108920,
 abstract = {This paper discusses the problem of reconstructing the state of a linear time invariant system when some of its actuators and sensors are compromised by an adversarial agent. In the model considered in this paper, the adversarial agent attacks an input (output) by manipulating its value arbitrarily, i.e., we impose no constraints (statistical or otherwise) on how control commands (sensor measurements) are changed by the adversary other than a bound on the number of attacked actuators and sensors In the first part of this paper, we introduce the notion of sparse strong observability and we show that is a necessary and sufficient condition for correctly reconstructing the state despite the considered attacks. In the second half of this work, we propose an observer to harness the complexity of this intrinsically combinatorial problem, by leveraging satisfiability modulo theory solving. Numerical simulations illustrate the effectiveness and scalability of our observer.},
 author = {Mehrdad Showkatbakhsh and Yasser Shoukry and Suhas N. Diggavi and Paulo Tabuada},
 doi = {https://doi.org/10.1016/j.automatica.2020.108920},
 issn = {0005-1098},
 journal = {Automatica},
 keywords = {Cyber–physical security, State reconstruction, Security monitoring},
 pages = {108920},
 tags = {journal,CPS},
 title = {Securing state reconstruction under sensor and actuator attacks: Theory and design},
 type = {2},
 url_link = {http://www.sciencedirect.com/science/article/pii/S0005109820301187},
 volume = {116},
 year = {2020}
}

This paper discusses the problem of reconstructing the state of a linear time invariant system when some of its actuators and sensors are compromised by an adversarial agent. In the model considered in this paper, the adversarial agent attacks an input (output) by manipulating its value arbitrarily, i.e., we impose no constraints (statistical or otherwise) on how control commands (sensor measurements) are changed by the adversary other than a bound on the number of attacked actuators and sensors In the first part of this paper, we introduce the notion of sparse strong observability and we show that is a necessary and sufficient condition for correctly reconstructing the state despite the considered attacks. In the second half of this work, we propose an observer to harness the complexity of this intrinsically combinatorial problem, by leveraging satisfiability modulo theory solving. Numerical simulations illustrate the effectiveness and scalability of our observer.

@article{diggavi2020coding,
 author = {Diggavi, Suhas and Tabuada, Paulo},
 doi = {10.1002/9781119593386.ch15},
 journal = {Modeling and Design of Secure Internet of Things},
 pages = {357--369},
 publisher = {Wiley Online Library},
 tags = {BookChap,CPS},
 title = {A Coding Theoretic View of Secure State Reconstruction},
 type = {3},
 year = {2020}
}

@article{hachem2020caching,
 author = {Hachem, J. and Niesen, U. and Diggavi, S.},
 editor = {Poor, Vince and Chen, Wei},
 journal = {Edge Caching for Mobile Networks},
 publisher = {Institute of Engineering and Technology (IET) Press},
 tags = {BookChap,CCWN,IT,ANIT,WiNetnew,SCS},
 title = {Caching for Interference Networks: A Separation Architecture},
 type = {3},
 year = {2020}
}

@article{hachem2020coded,
 author = {Hachem, Jad and Karamchandani, Nikhil and Diggavi, Suhas and Moharir, Sharayu},
 editor = {Vu, Thang and Chatzinotas, Symeon and Bastug, Ejder and Quek, Tony Q. S.},
 journal = {Wireless Edge Caching: Modeling, Analysis and Optimization},
 publishers = {Cambridge University Press},
 tags = {BookChap,CCWN,IT},
 title = {Coded Caching for Heterogeneous Wireless Networks},
 type = {3},
 year = {2020}
}

@inproceedings{chang2020wireless,
 author = {Chang, Kenneth and Raymondi, Nathaniel and Sabharwal, Ashutosh and Diggavi, Suhas},
 booktitle = {2020 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {1224--1229},
 tags = {conf,WiNetnew,WiIm},
 title = {" Wireless Paint": Code Design for 3D Orientation Estimation with Backscatter Arrays},
 type = {4},
 year = {2020}
}

@inproceedings{data2020byzantine,
 abstract = {We study stochastic gradient descent (SGD) in the master-worker architecture under Byzantine attacks. Building upon the recent advances in algorithmic high-dimensional robust statistics, in each SGD iteration, master employs a non-trivial decoding to estimate the true gradient from the unbiased stochastic gradients received from workers, some of which may be corrupt. We provide convergence analyses for both strongly-convex and non-convex smooth objectives under standard SGD assumptions. We can control the approximation error of our solution in both these settings by the mini-batch size of stochastic gradients; and we can make the approximation error as small as we want, provided that workers use a sufficiently large mini-batch size. Our algorithm can tolerate less than 1/3 fraction of Byzantine workers. It can approximately find the optimal parameters in the strongly-convex setting exponentially fast, and reaches to an approximate stationary point in the non-convex setting with linear speed, i.e., with a rate of 1/T, thus, matching the convergence rates of vanilla SGD in the Byzantine-free setting.},
 author = {Data, Deepesh and Diggavi, Suhas},
 booktitle = {2020 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {2628--2633},
 tags = {conf,SDL,DML},
 title = {On Byzantine-Resilient High-Dimensional Stochastic Gradient Descent},
 type = {4},
 doi = {10.1109/ISIT44484.2020.9174363},
 year = {2020}
}

We study stochastic gradient descent (SGD) in the master-worker architecture under Byzantine attacks. Building upon the recent advances in algorithmic high-dimensional robust statistics, in each SGD iteration, master employs a non-trivial decoding to estimate the true gradient from the unbiased stochastic gradients received from workers, some of which may be corrupt. We provide convergence analyses for both strongly-convex and non-convex smooth objectives under standard SGD assumptions. We can control the approximation error of our solution in both these settings by the mini-batch size of stochastic gradients; and we can make the approximation error as small as we want, provided that workers use a sufficiently large mini-batch size. Our algorithm can tolerate less than 1/3 fraction of Byzantine workers. It can approximately find the optimal parameters in the strongly-convex setting exponentially fast, and reaches to an approximate stationary point in the non-convex setting with linear speed, i.e., with a rate of 1/T, thus, matching the convergence rates of vanilla SGD in the Byzantine-free setting.

@inproceedings{girgis2020hiding,
 abstract = {In this paper, we study an estimation problem under the local differential privacy (LDP) framework: There is an ordered list of d values (e.g., real numbers); a set of n users, where each user observes an element from this list and each value in the list is observed by at least one user; and an untrusted server, who wants to estimate the values that the users possess, without learning (in the sense of LDP) the actual value that each user has and its corresponding index in the list. Towards this, we propose two LDP estimation schemes: The first one is under the assumption that the server knows the number of users that observe each value; and the second one is for the general scenario, in which the server does not have this prior information. We show that the minimax risk decreases with the total number of users under a very mild condition on the number of users observing each value.},
 author = {Girgis, Antonious M and Data, Deepesh and Diggavi, Suhas},
 booktitle = {2020 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {914--919},
 tags = {conf,DML,PDL},
 title = {Hiding identities: Estimation under local differential privacy},
 type = {4},
 doi = {10.1109/ISIT44484.2020.9174332},
 year = {2020}
}

In this paper, we study an estimation problem under the local differential privacy (LDP) framework: There is an ordered list of d values (e.g., real numbers); a set of n users, where each user observes an element from this list and each value in the list is observed by at least one user; and an untrusted server, who wants to estimate the values that the users possess, without learning (in the sense of LDP) the actual value that each user has and its corresponding index in the list. Towards this, we propose two LDP estimation schemes: The first one is under the assumption that the server knows the number of users that observe each value; and the second one is for the general scenario, in which the server does not have this prior information. We show that the minimax risk decreases with the total number of users under a very mild condition on the number of users observing each value.

@article{girgis2020shuffled,
 abstract = {We consider a distributed empirical risk minimization (ERM) optimization problem with communication efficiency and privacy requirements, motivated by the federated learning (FL) framework. Unique challenges to the traditional ERM problem in the context of FL include (i) need to provide privacy guarantees on clients' data, (ii) compress the communication between clients and the server, since clients might have low-bandwidth links, (iii) work with a dynamic client population at each round of communication between the server and the clients, as a small fraction of clients are sampled at each round. To address these challenges we develop (optimal) communication-efficient schemes for private mean estimation for several ℓp spaces, enabling efficient gradient aggregation for each iteration of the optimization solution of the ERM. We also provide lower and upper bounds for mean estimation with privacy and communication constraints for arbitrary ℓp spaces. To get the overall communication, privacy, and optimization performance operation point, we combine this with privacy amplification opportunities inherent to this setup. Our solution takes advantage of the inherent privacy amplification provided by client sampling and data sampling at each client (through Stochastic Gradient Descent) as well as the recently developed privacy framework using anonymization, which effectively presents to the server responses that are randomly shuffled with respect to the clients. Putting these together, we demonstrate that one can get the same privacy, optimization-performance operating point developed in recent methods that use full-precision communication, but at a much lower communication cost, i.e., effectively getting communication efficiency for "free".},
 author = {Girgis, Antonious M and Data, Deepesh and Diggavi, Suhas and Kairouz, Peter and Suresh, Ananda Theertha},
 journal = {NeurIPS Privacy-preserving Machine Learning workshop (PPML)},
 tags = {conf,DML,PDL},
 title = {Shuffled Model of Federated Learning: Privacy, Communication and Accuracy Trade-offs},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/2008.07180},
 year = {2020}
}

We consider a distributed empirical risk minimization (ERM) optimization problem with communication efficiency and privacy requirements, motivated by the federated learning (FL) framework. Unique challenges to the traditional ERM problem in the context of FL include (i) need to provide privacy guarantees on clients' data, (ii) compress the communication between clients and the server, since clients might have low-bandwidth links, (iii) work with a dynamic client population at each round of communication between the server and the clients, as a small fraction of clients are sampled at each round. To address these challenges we develop (optimal) communication-efficient schemes for private mean estimation for several ℓp spaces, enabling efficient gradient aggregation for each iteration of the optimization solution of the ERM. We also provide lower and upper bounds for mean estimation with privacy and communication constraints for arbitrary ℓp spaces. To get the overall communication, privacy, and optimization performance operation point, we combine this with privacy amplification opportunities inherent to this setup. Our solution takes advantage of the inherent privacy amplification provided by client sampling and data sampling at each client (through Stochastic Gradient Descent) as well as the recently developed privacy framework using anonymization, which effectively presents to the server responses that are randomly shuffled with respect to the clients. Putting these together, we demonstrate that one can get the same privacy, optimization-performance operating point developed in recent methods that use full-precision communication, but at a much lower communication cost, i.e., effectively getting communication efficiency for "free".

@article{singh2019sparq,
 abstract = {In this paper, we propose and analyze SPARQ-SGD, which is an event-triggered and compressed algorithm for decentralized training of large-scale machine learning models. Each node can locally compute a condition (event) which triggers a communication where quantized and sparsified local model parameters are sent. In SPARQ-SGD each node takes at least a fixed number (H) of local gradient steps and then checks if the model parameters have significantly changed compared to its last update; it communicates further compressed model parameters only when there is a significant change, as specified by a (design) criterion. We prove that the SPARQ-SGD converges as O(1nT) and O(1nT√) in the strongly-convex and non-convex settings, respectively, demonstrating that such aggressive compression, including event-triggered communication, model sparsification and quantization does not affect the overall convergence rate as compared to uncompressed decentralized training; thereby theoretically yielding communication efficiency for "free". We evaluate SPARQ-SGD over real datasets to demonstrate significant amount of savings in communication over the state-of-the-art.},
 author = {Singh, Navjot and Data, Deepesh and George, Jemin and Diggavi, Suhas},
 journal = {2020 59th IEEE Conference on Decision and Control (CDC)},
 tags = {conf,CEDL,DML},
 title = {SPARQ-SGD: Event-Triggered and Compressed Communication in Decentralized Stochastic Optimization},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/1910.14280},
 year = {2020},
 pages={3449-3456},
 doi={10.1109/CDC42340.2020.9303828},
 ISSN={2576-2370},
}

In this paper, we propose and analyze SPARQ-SGD, which is an event-triggered and compressed algorithm for decentralized training of large-scale machine learning models. Each node can locally compute a condition (event) which triggers a communication where quantized and sparsified local model parameters are sent. In SPARQ-SGD each node takes at least a fixed number (H) of local gradient steps and then checks if the model parameters have significantly changed compared to its last update; it communicates further compressed model parameters only when there is a significant change, as specified by a (design) criterion. We prove that the SPARQ-SGD converges as O(1nT) and O(1nT√) in the strongly-convex and non-convex settings, respectively, demonstrating that such aggressive compression, including event-triggered communication, model sparsification and quantization does not affect the overall convergence rate as compared to uncompressed decentralized training; thereby theoretically yielding communication efficiency for "free". We evaluate SPARQ-SGD over real datasets to demonstrate significant amount of savings in communication over the state-of-the-art.

@inproceedings{srinivasavaradhan2020equivalence,
 abstract = {Maximum likelihood (ML) and symbolwise maximum aposteriori (MAP) estimation for discrete input sequences play a central role in a number of applications that arise in communications, information and coding theory. Many instances of these problems are proven to be intractable, for example through reduction to NP-complete integer optimization problems. In this work, we prove that the ML estimation of a discrete input sequence (with no assumptions on the encoder/channel used) is equivalent to the solution of a continuous non-convex optimization problem, and that this formulation is closely related to the computation of symbolwise MAP estimates. This equivalence is particularly useful in situations where a function we term the expected likelihood is efficiently computable. In such situations, we give a ML heuristic and show numerics for sequence estimation over the deletion channel.},
 author = {Srinivasavaradhan, Sundara Rajan and Diggavi, Suhas and Fragouli, Christina},
 booktitle = {2020 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {343--348},
 tags = {conf,BioInf,IT,NDS},
 title = {Equivalence of ml decoding to a continuous optimization problem},
 type = {4},
 doi = {10.1109/ISIT44484.2020.9174130},
 year = {2020}
}

Maximum likelihood (ML) and symbolwise maximum aposteriori (MAP) estimation for discrete input sequences play a central role in a number of applications that arise in communications, information and coding theory. Many instances of these problems are proven to be intractable, for example through reduction to NP-complete integer optimization problems. In this work, we prove that the ML estimation of a discrete input sequence (with no assumptions on the encoder/channel used) is equivalent to the solution of a continuous non-convex optimization problem, and that this formulation is closely related to the computation of symbolwise MAP estimates. This equivalence is particularly useful in situations where a function we term the expected likelihood is efficiently computable. In such situations, we give a ML heuristic and show numerics for sequence estimation over the deletion channel.

@article{karakus2019redundancy,
 abstract = {Performance of distributed optimization and learning systems is bottlenecked by "straggler" nodes and slow communication links, which significantly delay computation. We propose a distributed optimization framework where the dataset is "encoded" to have an over-complete representation with built-in redundancy, and the straggling nodes in the system are dynamically left out of the computation at every iteration, whose loss is compensated by the embedded redundancy. We show that oblivious application of several popular optimization algorithms on encoded data, including gradient descent, L-BFGS, proximal gradient under data parallelism, and coordinate descent under model parallelism, converge to either approximate or exact solutions of the original problem when stragglers are treated as erasures. These convergence results are deterministic, i.e., they establish sample path convergence for arbitrary sequences of delay patterns or distributions on the nodes, and are independent of the tail behavior of the delay distribution. We demonstrate that equiangular tight frames have desirable properties as encoding matrices, and propose efficient mechanisms for encoding large-scale data. We implement the proposed technique on Amazon EC2 clusters, and demonstrate its performance over several learning problems, including matrix factorization, LASSO, ridge regression and logistic regression, and compare the proposed method with uncoded, asynchronous, and data replication strategies.},
 author = {Karakus, Can and Sun, Yifan and Diggavi, Suhas N and Yin, Wotao},
 journal = {J. Mach. Learn. Res.},
 number = {72},
 pages = {1--47},
 tags = {journal,DML,EncOpt},
 title = {Redundancy Techniques for Straggler Mitigation in Distributed Optimization and Learning.},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1803.05397},
 volume = {20},
 year = {2019}
}

Performance of distributed optimization and learning systems is bottlenecked by "straggler" nodes and slow communication links, which significantly delay computation. We propose a distributed optimization framework where the dataset is "encoded" to have an over-complete representation with built-in redundancy, and the straggling nodes in the system are dynamically left out of the computation at every iteration, whose loss is compensated by the embedded redundancy. We show that oblivious application of several popular optimization algorithms on encoded data, including gradient descent, L-BFGS, proximal gradient under data parallelism, and coordinate descent under model parallelism, converge to either approximate or exact solutions of the original problem when stragglers are treated as erasures. These convergence results are deterministic, i.e., they establish sample path convergence for arbitrary sequences of delay patterns or distributions on the nodes, and are independent of the tail behavior of the delay distribution. We demonstrate that equiangular tight frames have desirable properties as encoding matrices, and propose efficient mechanisms for encoding large-scale data. We implement the proposed technique on Amazon EC2 clusters, and demonstrate its performance over several learning problems, including matrix factorization, LASSO, ridge regression and logistic regression, and compare the proposed method with uncoded, asynchronous, and data replication strategies.

@article{karmoose2019using,
 abstract = {The fact that millimeter-wave (mmWave) communication needs to be directional is usually perceived as a challenge; in this letter, we argue that it enables efficient secret key sharing that is unconditionally secure from passive eavesdroppers by building on packet erasures. We showcase the potential of our approach in two setups: 5G networks and vehicle platooning. In both cases, with the right choice of parameters, we can potentially establish keys in the order of tenths of Mb/s. These first results are based on some simplifying assumptions, yet we believe that they give incentives to further explore such techniques.},
 author = {Karmoose, Mohammed and Fragouli, Christina and Diggavi, Suhas and Misoczki, Rafael and Yang, Lily L and Zhang, Zhenliang},
 journal = {IEEE Communications Letters},
 number = {6},
 pages = {1077--1080},
 publisher = {IEEE},
 tags = {journal,WNS,WiNetnew},
 title = {Using mm-waves for secret key establishment},
 type = {2},
 doi = {10.1109/LCOMM.2019.2909918},
 volume = {23},
 year = {2019}
}

The fact that millimeter-wave (mmWave) communication needs to be directional is usually perceived as a challenge; in this letter, we argue that it enables efficient secret key sharing that is unconditionally secure from passive eavesdroppers by building on packet erasures. We showcase the potential of our approach in two setups: 5G networks and vehicle platooning. In both cases, with the right choice of parameters, we can potentially establish keys in the order of tenths of Mb/s. These first results are based on some simplifying assumptions, yet we believe that they give incentives to further explore such techniques.

@article{wang2019slate,
 abstract = {Lightweight cryptography has become more and more important in recent years because of the rise of the Internet of Things (IoT) and usage of smart mobile devices. In this paper, we propose a novel secure lightweight entity authentication hardware primitive called SLATE, where its area is about 50% to more than 3X smaller than existing lightweight ciphers and strong physical unclonable functions (PUFs), respectively. Even though the authentication of SLATE is done through challenge response pair (CRP) verification similar to strong PUFs, the source of the key for SLATE must be coming from any existing secret key storage used for any ciphers. A main advantage of SLATE over most existing strong PUFs being an entity authentication primitive is that SLATE is resistant to known attacks to strong PUFs or logic obfuscations, such as model building attacks and Boolean satisfiability (SAT) attacks. Furthermore, we show that the implementation cost of SLATE with a 176-bit key and 244 CRPs is only 663 gate equivalents (GEs). Compared with lightweight ciphers and existing secure strong PUFs, we show that SLATE is a practical security primitive for resource constrained systems for its extremely small footprint and security. Finally, we show that SLATE is information theoretically secure when valid CRPs are communicated through insecure channels.},
 author = {Wang, Wei-Che and Yona, Yair and Wu, Yizhang and Diggavi, Suhas N and Gupta, Puneet},
 journal = {IEEE Transactions on Information Forensics and Security},
 pages = {276--285},
 publisher = {IEEE},
 tags = {journal,HS},
 title = {SLATE: A Secure Lightweight Entity Authentication Hardware Primitive},
 type = {2},
 doi = {10.1109/TIFS.2019.2919393},
 volume = {15},
 year = {2019}
}

Lightweight cryptography has become more and more important in recent years because of the rise of the Internet of Things (IoT) and usage of smart mobile devices. In this paper, we propose a novel secure lightweight entity authentication hardware primitive called SLATE, where its area is about 50% to more than 3X smaller than existing lightweight ciphers and strong physical unclonable functions (PUFs), respectively. Even though the authentication of SLATE is done through challenge response pair (CRP) verification similar to strong PUFs, the source of the key for SLATE must be coming from any existing secret key storage used for any ciphers. A main advantage of SLATE over most existing strong PUFs being an entity authentication primitive is that SLATE is resistant to known attacks to strong PUFs or logic obfuscations, such as model building attacks and Boolean satisfiability (SAT) attacks. Furthermore, we show that the implementation cost of SLATE with a 176-bit key and 244 CRPs is only 663 gate equivalents (GEs). Compared with lightweight ciphers and existing secure strong PUFs, we show that SLATE is a practical security primitive for resource constrained systems for its extremely small footprint and security. Finally, we show that SLATE is information theoretically secure when valid CRPs are communicated through insecure channels.

@inproceedings{basu2019qsparse,
 abstract = {Communication bottleneck has been identified as a significant issue in distributed optimization of large-scale learning models. Recently, several approaches to mitigate this problem have been proposed, including different forms of gradient compression or computing local models and mixing them iteratively. In this paper, we propose \emph{Qsparse-local-SGD} algorithm, which combines aggressive sparsification with quantization and local computation along with error compensation, by keeping track of the difference between the true and compressed gradients. We propose both synchronous and asynchronous implementations of \emph{Qsparse-local-SGD}. We analyze convergence for \emph{Qsparse-local-SGD} in the \emph{distributed} setting for smooth non-convex and convex objective functions. We demonstrate that \emph{Qsparse-local-SGD} converges at the same rate as vanilla distributed SGD for many important classes of sparsifiers and quantizers. We use \emph{Qsparse-local-SGD} to train ResNet-50 on ImageNet and show that it results in significant savings over the state-of-the-art, in the number of bits transmitted to reach target accuracy.},
 author = {Basu, Debraj and Data, Deepesh and Karakus, Can and Diggavi, Suhas},
 booktitle = {Advances in Neural Information Processing Systems},
 pages = {14695--14706},
 tags = {conf,CEDL,DML},
 title = {Qsparse-local-SGD: Distributed SGD with quantization, sparsification and local computations},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/1906.02367},
 year = {2019}
}

Communication bottleneck has been identified as a significant issue in distributed optimization of large-scale learning models. Recently, several approaches to mitigate this problem have been proposed, including different forms of gradient compression or computing local models and mixing them iteratively. In this paper, we propose \emphQsparse-local-SGD algorithm, which combines aggressive sparsification with quantization and local computation along with error compensation, by keeping track of the difference between the true and compressed gradients. We propose both synchronous and asynchronous implementations of \emphQsparse-local-SGD. We analyze convergence for \emphQsparse-local-SGD in the \emphdistributed setting for smooth non-convex and convex objective functions. We demonstrate that \emphQsparse-local-SGD converges at the same rate as vanilla distributed SGD for many important classes of sparsifiers and quantizers. We use \emphQsparse-local-SGD to train ResNet-50 on ImageNet and show that it results in significant savings over the state-of-the-art, in the number of bits transmitted to reach target accuracy.

@inproceedings{data2019byzantine,
 abstract = {We study distributed coordinate descent (CD) in the master-worker architecture under adversarial attacks, where the data is partitioned (across the parameter space) and distributed among m worker nodes (t of which can be maliciously corrupt), which update some coordinates of their part of the parameter vector, in parallel and iteratively, using CD updates, with the help of the master. We propose a method based on data encoding and real error correction to combat the adversary. Our method can tolerate up to ⌈m-1/2⌉ corrupt nodes, which is information-theoretically optimal. Our design gives a trade-off between the resiliency t, the required redundancy, and the computation at master and worker nodes. For example, with constant overhead in the storage and computational complexity over that required by the plain distributed CD, we can tolerate up to m/3 corrupt nodes. We design a sparse encoding scheme, which yields low encoding complexity.},
 author = {Data, Deepesh and Diggavi, Suhas},
 booktitle = {2019 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {2724--2728},
 tags = {conf,SDL,DML},
 title = {Byzantine-Tolerant Distributed Coordinate Descent},
 type = {4},
 doi = {10.1109/ISIT.2019.8849217},
 year = {2019}
}

We study distributed coordinate descent (CD) in the master-worker architecture under adversarial attacks, where the data is partitioned (across the parameter space) and distributed among m worker nodes (t of which can be maliciously corrupt), which update some coordinates of their part of the parameter vector, in parallel and iteratively, using CD updates, with the help of the master. We propose a method based on data encoding and real error correction to combat the adversary. Our method can tolerate up to ⌈m-1/2⌉ corrupt nodes, which is information-theoretically optimal. Our design gives a trade-off between the resiliency t, the required redundancy, and the computation at master and worker nodes. For example, with constant overhead in the storage and computational complexity over that required by the plain distributed CD, we can tolerate up to m/3 corrupt nodes. We design a sparse encoding scheme, which yields low encoding complexity.

@inproceedings{data2019data,
 abstract = {We consider distributed gradient computation, where both data and computation are distributed among m worker machines, t of which can be Byzantine adversaries, and a designated (master) node computes the model/parameter vector for generalized linear models, iteratively, using proximal gradient descent (PGD), of which gradient descent (GD) is a special case. The Byzantine adversaries can (collaboratively) deviate arbitrarily from their gradient computation. To solve this, we propose a method based on data encoding and (real) error correction to combat the adversarial behavior. We can tolerate up to t ≤ [m-1/2] corrupt worker nodes, which is information-theoretically optimal. Our method does not assume any probability distribution on the data. We develop a sparse encoding scheme which enables computationally efficient data encoding. We demonstrate a trade-off between the number of adversaries tolerated and the resource requirement (storage and computational complexity). As an example, our scheme incurs a constant overhead (storage and computational complexity) over that required by the distributed PGD algorithm, without adversaries, for t ≤ m/3 . Our encoding works as efficiently in the streaming data etting as it does in the},
 author = {Data, Deepesh and Song, Linqi and Diggavi, Suhas},
 booktitle = {2019 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {2719--2723},
 tags = {conf,SDL,DML},
 title = {Data encoding methods for byzantine-resilient distributed optimization},
 type = {4},
 doi = {10.1109/ISIT.2019.8849857},
 year = {2019}
}

We consider distributed gradient computation, where both data and computation are distributed among m worker machines, t of which can be Byzantine adversaries, and a designated (master) node computes the model/parameter vector for generalized linear models, iteratively, using proximal gradient descent (PGD), of which gradient descent (GD) is a special case. The Byzantine adversaries can (collaboratively) deviate arbitrarily from their gradient computation. To solve this, we propose a method based on data encoding and (real) error correction to combat the adversarial behavior. We can tolerate up to t ≤ [m-1/2] corrupt worker nodes, which is information-theoretically optimal. Our method does not assume any probability distribution on the data. We develop a sparse encoding scheme which enables computationally efficient data encoding. We demonstrate a trade-off between the number of adversaries tolerated and the resource requirement (storage and computational complexity). As an example, our scheme incurs a constant overhead (storage and computational complexity) over that required by the distributed PGD algorithm, without adversaries, for t ≤ m/3 . Our encoding works as efficiently in the streaming data etting as it does in the

@inproceedings{ezzeldin2019quantizing,
 abstract = {In many machine learning applications, once we have learned a classifier, in order to apply it, we may still need to gather features from distributed sensors over communication constrained channels. In this paper, we propose a polynomial complexity algorithm for feature quantization tailored to minimizing the classification error of a linear classifier. Our scheme produces scalar quantizers that are well-tailored to delay-sensitive applications, operates on the same training data used to learn the classifier, and allows each distributed sensor to operate independently of each other. Numerical evaluation indicates up to 65% benefits over alternative approaches. Additionally, we provide an example where, jointly designing the linear classifier and the quantization scheme, can outperform sequential designs.},
 author = {Ezzeldin, Yahya H and Fragouli, Christina and Diggavi, Suhas},
 booktitle = {2019 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {912--916},
 tags = {conf,DML,CEDL},
 title = {Quantizing Signals for Linear Classification},
 type = {4},
 doi = {10.1109/ISIT.2019.8849589},
 year = {2019}
}

In many machine learning applications, once we have learned a classifier, in order to apply it, we may still need to gather features from distributed sensors over communication constrained channels. In this paper, we propose a polynomial complexity algorithm for feature quantization tailored to minimizing the classification error of a linear classifier. Our scheme produces scalar quantizers that are well-tailored to delay-sensitive applications, operates on the same training data used to learn the classifier, and allows each distributed sensor to operate independently of each other. Numerical evaluation indicates up to 65% benefits over alternative approaches. Additionally, we provide an example where, jointly designing the linear classifier and the quantization scheme, can outperform sequential designs.

@inproceedings{shah2019coded,
 abstract = {In the coded caching problem a file server with N files is connected through an error-free broadcast link to K users, each equipped with a local cache. The placement and delivery phases can be jointly optimized to create simultaneous multicasting opportunities among users that request different files. We examine the model where we have multiple user classes and users belonging to different user classes have different demand popularities. We develop placement schemes based on either the global popularity of the files or the local popularity of the files and for each of these placement schemes, we examine delivery schemes based on either the user perspective or the file perspective. We analyze the server transmission rates for these schemes and identify parameter regimes where schemes working with local content popularity profiles do better than those utilizing only the global popularity.},
 author = {Shah, Abhin and Karamchandani, Nikhil and Diggavi, Suhas},
 booktitle = {2019 16th Canadian Workshop on Information Theory (CWIT)},
 organization = {IEEE},
 pages = {1--6},
 tags = {conf,CCWN},
 title = {Coded Caching: Global vs Local Content Popularity},
 type = {4},
 doi = {10.1109/CWIT.2019.8929932},
 year = {2019}
}

In the coded caching problem a file server with N files is connected through an error-free broadcast link to K users, each equipped with a local cache. The placement and delivery phases can be jointly optimized to create simultaneous multicasting opportunities among users that request different files. We examine the model where we have multiple user classes and users belonging to different user classes have different demand popularities. We develop placement schemes based on either the global popularity of the files or the local popularity of the files and for each of these placement schemes, we examine delivery schemes based on either the user perspective or the file perspective. We analyze the server transmission rates for these schemes and identify parameter regimes where schemes working with local content popularity profiles do better than those utilizing only the global popularity.

@inproceedings{srinivasavaradhan2019symbolwise,
 abstract = {We consider the problem of reconstructing a sequence from fixed number of deleted versions of itself (also called traces). The problem is motivated from recent developments in de novo DNA sequencing technologies. The main contribution of this work is to provide a polynomial time algorithm for symbolwise MAP decoding with multiple traces. The algorithm leverages a dynamic program on the edit graph. We also develop a heuristic with reduced time complexity using similar ideas and provide preliminary numerical evaluations.},
 author = {Srinivasavaradhan, Sundara R and Du, Michelle and Diggavi, Suhas and Fragouli, Christina},
 booktitle = {2019 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {181--185},
 tags = {conf,BioInf,IT,NDS},
 title = {Symbolwise map for multiple deletion channels},
 type = {4},
 doi = {10.1109/ISIT.2019.8849567},
 year = {2019}
}

We consider the problem of reconstructing a sequence from fixed number of deleted versions of itself (also called traces). The problem is motivated from recent developments in de novo DNA sequencing technologies. The main contribution of this work is to provide a polynomial time algorithm for symbolwise MAP decoding with multiple traces. The algorithm leverages a dynamic program on the edit graph. We also develop a heuristic with reduced time complexity using similar ideas and provide preliminary numerical evaluations.

@article{sebastian2018generalized,
 abstract = {We study the generalized degrees of freedom (gDoF) of the block-fading noncoherent 2-user interference channel (IC) with a coherence time of T symbol durations and symmetric fading statistics. We demonstrate that a natural training-based scheme, to operate the noncoherent IC, is suboptimal in several regimes. As an alternate scheme, we propose a new noncoherent rate-splitting scheme. We also consider treating interference-as-noise (TIN) scheme and a time division multiplexing (TDM) scheme. We observe that a standard training-based scheme for IC is outperformed by one of these schemes in several regimes: our results demonstrate that for low average interference-to-noise ratio (INR), TIN is best; for high INR, TDM and the noncoherent rate-splitting give better performance. We also study the noncoherent IC with feedback and propose a noncoherent rate-splitting scheme. Again for the feedback case as well, our results demonstrate that a natural training-based scheme can be outperformed by other schemes.},
 author = {Sebastian, Joyson and Diggavi, Suhas},
 journal = {arXiv preprint arXiv:1812.03579},
 tags = {journalSub,IT,ANIT,WiNetnew,NCWN},
 title = {On the Generalized Degrees of Freedom of Noncoherent Interference Channel},
 type = {1},
 url_arxiv = {https://arxiv.org/abs/1812.03579},
 year = {2018}
}

We study the generalized degrees of freedom (gDoF) of the block-fading noncoherent 2-user interference channel (IC) with a coherence time of T symbol durations and symmetric fading statistics. We demonstrate that a natural training-based scheme, to operate the noncoherent IC, is suboptimal in several regimes. As an alternate scheme, we propose a new noncoherent rate-splitting scheme. We also consider treating interference-as-noise (TIN) scheme and a time division multiplexing (TDM) scheme. We observe that a standard training-based scheme for IC is outperformed by one of these schemes in several regimes: our results demonstrate that for low average interference-to-noise ratio (INR), TIN is best; for high INR, TDM and the noncoherent rate-splitting give better performance. We also study the noncoherent IC with feedback and propose a noncoherent rate-splitting scheme. Again for the feedback case as well, our results demonstrate that a natural training-based scheme can be outperformed by other schemes.

@article{8301564,
 abstract = {Nanopore sequencing is an emerging new technology for sequencing Deoxyribonucleic acid (DNA), which can read long fragments of DNA (~50000 bases), in contrast to most current short-read sequencing technologies which can only read hundreds of bases. While nanopore sequencers can acquire long reads, the high error rates (20\%-30\%) pose a technical challenge. In a nanopore sequencer, a DNA is migrated through a nanopore, and current variations are measured. The DNA sequence is inferred from this observed current pattern using an algorithm called a base-caller. In this paper, we propose a mathematical model for the “channel” from the input DNA sequence to the observed current, and calculate bounds on the information extraction capacity of the nanopore sequencer. This model incorporates impairments, such as (non-linear) intersymbol interference, deletions, and random response. These information bounds have two-fold application: 1) The decoding rate with a uniform input distribution can be used to calculate the average size of the plausible list of DNA sequences given an observed current trace. This bound can be used to benchmark existing base-calling algorithms, as well as serving a performance objective to design better nanopores. 2) When the nanopore sequencer is used as a reader in a DNA storage system, the storage capacity is quantified by our bounds.},
 author = {W. {Mao} and S. N. {Diggavi} and S. {Kannan}},
 doi = {10.1109/TIT.2018.2809001},
 issn = {1557-9654},
 journal = {IEEE Transactions on Information Theory},
 keywords = {biology computing;DNA;genomics;intersymbol interference;molecular biophysics;nanobiotechnology;nanopore sequencing;sequencing Deoxyribonucleic acid;sequencing technologies;nanopore sequencer;input DNA sequences;base-calling algorithms;DNA storage system;storage capacity;DNA;Sequential analysis;Decoding;Nanobioscience;Current measurement;Reliability;Mathematical model;Deoxyribonucleic acid (DNA) sequencing;bioinformatics;base calling;channel with synchronization errors;deletion channel;finite state channels},
 month = {April},
 number = {4},
 pages = {3216-3236},
 tags = {journal,IT,BioInf,NDS},
 title = {Models and Information-Theoretic Bounds for Nanopore Sequencing},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1705.11154},
 volume = {64},
 year = {2018}
}

Nanopore sequencing is an emerging new technology for sequencing Deoxyribonucleic acid (DNA), which can read long fragments of DNA ( 50000 bases), in contrast to most current short-read sequencing technologies which can only read hundreds of bases. While nanopore sequencers can acquire long reads, the high error rates (20%-30%) pose a technical challenge. In a nanopore sequencer, a DNA is migrated through a nanopore, and current variations are measured. The DNA sequence is inferred from this observed current pattern using an algorithm called a base-caller. In this paper, we propose a mathematical model for the “channel” from the input DNA sequence to the observed current, and calculate bounds on the information extraction capacity of the nanopore sequencer. This model incorporates impairments, such as (non-linear) intersymbol interference, deletions, and random response. These information bounds have two-fold application: 1) The decoding rate with a uniform input distribution can be used to calculate the average size of the plausible list of DNA sequences given an observed current trace. This bound can be used to benchmark existing base-calling algorithms, as well as serving a performance objective to design better nanopores. 2) When the nanopore sequencer is used as a reader in a DNA storage system, the storage capacity is quantified by our bounds.

@article{8335768,
 abstract = {This letter initiates the study of energy-efficiency gains provided by caching. We focus on the cache-aided Gaussian interference channel in the low-SNR regime. We propose a strategy that creates content overlaps at the transmitter caches to allow for co-operation between the transmitters. This co-operation yields a beamforming gain, which has to be traded off against a multicasting gain. We evaluate the performance of this strategy and show its approximate optimality in both the single-receiver case and the single-transmitter case.},
 author = {J. {Hachem} and U. {Niesen} and S. {Diggavi}},
 doi = {10.1109/LCOMM.2018.2822694},
 issn = {1558-2558},
 journal = {IEEE Communications Letters},
 keywords = {array signal processing;cache storage;channel capacity;energy conservation;Gaussian channels;multicast communication;radio networks;radiofrequency interference;telecommunication power management;wireless channels;energy-efficiency gains;cache-aided Gaussian interference channel;low-SNR regime;transmitter caches;co-operation yields;beamforming gain;multicasting gain;Transmitters;Receivers;Multicast communication;Array signal processing;Libraries;Interference channels;Network coding;wireless networks;content distribution networks},
 month = {July},
 number = {7},
 pages = {1434-1437},
 tags = {journal,CCWN,IT,ANIT,WiNetnew,SCS},
 title = {Energy-Efficiency Gains of Caching for Interference Channels},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1808.00653},
 volume = {22},
 year = {2018}
}

This letter initiates the study of energy-efficiency gains provided by caching. We focus on the cache-aided Gaussian interference channel in the low-SNR regime. We propose a strategy that creates content overlaps at the transmitter caches to allow for co-operation between the transmitters. This co-operation yields a beamforming gain, which has to be traded off against a multicasting gain. We evaluate the performance of this strategy and show its approximate optimality in both the single-receiver case and the single-transmitter case.

@article{8429509,
 abstract = {We develop a characterization of fading models, which assigns a number called logarithmic Jensen's gap to a given fading model. We show that as a consequence of a finite logarithmic Jensen's gap, an approximate capacity region can be obtained for fast fading interference channels (FF-ICs) for several scenarios. We illustrate three instances where a constant capacity gap can be obtained as a function of the logarithmic Jensen's gap. First, for an FF-IC with neither feedback nor instantaneous channel state information at transmitter (CSIT), if the fading distribution has finite logarithmic Jensen's gap, we show that a rate-splitting scheme based on the average interference-to-noise ratio can achieve its approximate capacity. Second, we show that a similar scheme can achieve the approximate capacity of FF-IC with feedback and delayed CSIT, if the fading distribution has finite logarithmic Jensen's gap. Third, when this condition holds, we show that point-to-point codes can achieve approximate capacity for a class of FF-ICs with feedback. We prove that the logarithmic Jensen's gap is finite for common fading models, including Rayleigh and Nakagami fading, thereby obtaining the approximate capacity region of FF-IC with these fading models.},
 author = {J. {Sebastian} and C. {Karakus} and S. {Diggavi}},
 doi = {10.1109/TCOMM.2018.2864266},
 issn = {1558-0857},
 journal = {IEEE Transactions on Communications},
 keywords = {approximation theory;channel capacity;diversity reception;fading channels;Gaussian channels;MIMO communication;Nakagami channels;radio transmitters;radiofrequency interference;Rayleigh channels;fast fading interference channels;finite logarithmic Jensen's gap;approximate capacity region;FF-IC;constant capacity gap;instantaneous channel state information;fading distribution;common fading models;Rayleigh channels;Receivers;Transmitters;Integrated circuit modeling;Interference;Interference channels;fast fading;capacity region;rate-splitting},
 month = {Dec},
 number = {12},
 pages = {6015-6027},
 tags = {journal,ANIT,WiNetnew,NCWN,WNIF},
 title = {Approximate Capacity of Fast Fading Interference Channels With no Instantaneous CSIT},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1706.03659},
 volume = {66},
 year = {2018}
}

We develop a characterization of fading models, which assigns a number called logarithmic Jensen's gap to a given fading model. We show that as a consequence of a finite logarithmic Jensen's gap, an approximate capacity region can be obtained for fast fading interference channels (FF-ICs) for several scenarios. We illustrate three instances where a constant capacity gap can be obtained as a function of the logarithmic Jensen's gap. First, for an FF-IC with neither feedback nor instantaneous channel state information at transmitter (CSIT), if the fading distribution has finite logarithmic Jensen's gap, we show that a rate-splitting scheme based on the average interference-to-noise ratio can achieve its approximate capacity. Second, we show that a similar scheme can achieve the approximate capacity of FF-IC with feedback and delayed CSIT, if the fading distribution has finite logarithmic Jensen's gap. Third, when this condition holds, we show that point-to-point codes can achieve approximate capacity for a class of FF-ICs with feedback. We prove that the logarithmic Jensen's gap is finite for common fading models, including Rayleigh and Nakagami fading, thereby obtaining the approximate capacity region of FF-IC with these fading models.

@article{abdelzaher2018toward,
 abstract = {The Internet of Battlefield Things (IoBT) might be one of the most expensive cyber-physical systems of the next decade, yet much research remains to develop its fundamental enablers. A challenge that distinguishes the IoBT from its civilian counterparts is resilience to a much larger spectrum of threats.},
 author = {Abdelzaher, Tarek and Ayanian, Nora and Basar, Tamer and Diggavi, Suhas and Diesner, Jana and Ganesan, Deepak and Govindan, Ramesh and Jha, Susmit and Lepoint, Tancrede and Marlin, Benjamin and others},
 journal = {Computer},
 number = {11},
 pages = {24--36},
 publisher = {IEEE},
 tags = {journal,CPS,DML},
 title = {Toward an internet of battlefield things: a resilience perspective},
 type = {2},
 doi = {10.1109/MC.2018.2876048},
 volume = {51},
 year = {2018}
}

The Internet of Battlefield Things (IoBT) might be one of the most expensive cyber-physical systems of the next decade, yet much research remains to develop its fundamental enablers. A challenge that distinguishes the IoBT from its civilian counterparts is resilience to a much larger spectrum of threats.

@article{DesignAnalysisPUFs2017,
 abstract = {The lack of stability is one of the major limitations that constrains PUF from being put in widespread practical use. In this paper, we propose a weak PUF and a strong PUF that are both completely stable with 0% intra-distance. These PUFs are called Locally Enhanced Defectivity (LED)PUF. The source of randomness of a LEDPUF is extracted from locally enhance defectivity without affecting other parts of the chip. A LEDPUF is a pure functional PUF that does not require any kinds of correction schemes as conventional parametric PUFs do. A weak LEDPUF is constructed by forming arrays of Directed Self Assembly (DSA) random connections is presented, and the strong LEDPUF is implemented by using the weak LEDPUF as the key of a keyed-hash message authentication code (HMAC). Our simulation and statistical results show that the entropy of the weak LEDPUF bits is close to ideal, and the inter-distances of both weak and strong LEDPUFs are about 50%, which means that these LEDPUFs are not only stable but also unique. We develop a new unified framework for evaluating the level of security of PUFs, based on password security, by using information theoretic tools of guesswork. The guesswork model allows to quantitatively compare, with a single unified metric, PUFs with varying levels of stability, bias and available side information. In addition, it generalizes other measures to evaluate the security level such as min-entropy and mutual information. We evaluate guesswork-based security of some measured SRAM and Ring Oscillator PUFs as an example and compare them with LEDPUF to show that stability has a more severe impact on the PUF security than biased responses. Furthermore, we find the guesswork of three new problems: Guesswork under probability of attack failure, the guesswork of idealized version of a message authentication code, and the guesswork of strong PUFs that are used for authentication.},
 author = {W. C. Wang and Y. Yona and S. Diggavi and P. Gupta},
 file = {:papers:designanalysispufs2017.pdf},
 journal = {IEEE Transactions on Information Forensics and Security},
 number = {4},
 pages = {978-992},
 tags = {journal,IT,Hardware},
 title = {Design and Analysis of Stability-Guaranteed PUFs},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1701.05637},
 volume = {13},
 year = {2018}
}

The lack of stability is one of the major limitations that constrains PUF from being put in widespread practical use. In this paper, we propose a weak PUF and a strong PUF that are both completely stable with 0% intra-distance. These PUFs are called Locally Enhanced Defectivity (LED)PUF. The source of randomness of a LEDPUF is extracted from locally enhance defectivity without affecting other parts of the chip. A LEDPUF is a pure functional PUF that does not require any kinds of correction schemes as conventional parametric PUFs do. A weak LEDPUF is constructed by forming arrays of Directed Self Assembly (DSA) random connections is presented, and the strong LEDPUF is implemented by using the weak LEDPUF as the key of a keyed-hash message authentication code (HMAC). Our simulation and statistical results show that the entropy of the weak LEDPUF bits is close to ideal, and the inter-distances of both weak and strong LEDPUFs are about 50%, which means that these LEDPUFs are not only stable but also unique. We develop a new unified framework for evaluating the level of security of PUFs, based on password security, by using information theoretic tools of guesswork. The guesswork model allows to quantitatively compare, with a single unified metric, PUFs with varying levels of stability, bias and available side information. In addition, it generalizes other measures to evaluate the security level such as min-entropy and mutual information. We evaluate guesswork-based security of some measured SRAM and Ring Oscillator PUFs as an example and compare them with LEDPUF to show that stability has a more severe impact on the PUF security than biased responses. Furthermore, we find the guesswork of three new problems: Guesswork under probability of attack failure, the guesswork of idealized version of a message authentication code, and the guesswork of strong PUFs that are used for authentication.

@article{hachem2018caching,
 abstract = {We study the caching problem when we are allowed to match each user to one of a subset of caches after its request is revealed. We focus on non-uniformly popular content, specifically when the file popularities obey a Zipf distribution. We study two extremal schemes: one focusing on coded server transmissions while ignoring matching capabilities and the other focusing on adaptive matching while ignoring potential coding opportunities. We derive the rates achieved by these schemes and characterize the regimes in which one outperforms the other. We also compare them to information-theoretic outer bounds and finally propose a hybrid scheme that generalizes ideas from the two schemes and performs at least as well as either of them in most memory regimes.},
 author = {Hachem, Jad and Karamchandani, Nikhil and Moharir, Sharayu and Diggavi, Suhas N},
 journal = {IEEE Journal on Selected Areas in Communications},
 number = {8},
 pages = {1831--1842},
 publisher = {IEEE},
 tags = {journal,CCWN,IT},
 title = {Caching with partial adaptive matching},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1705.03852},
 doi = {10.1109/JSAC.2018.2845018},
 volume = {36},
 year = {2018}
}

We study the caching problem when we are allowed to match each user to one of a subset of caches after its request is revealed. We focus on non-uniformly popular content, specifically when the file popularities obey a Zipf distribution. We study two extremal schemes: one focusing on coded server transmissions while ignoring matching capabilities and the other focusing on adaptive matching while ignoring potential coding opportunities. We derive the rates achieved by these schemes and characterize the regimes in which one outperforms the other. We also compare them to information-theoretic outer bounds and finally propose a hybrid scheme that generalizes ideas from the two schemes and performs at least as well as either of them in most memory regimes.

@article{hachem2018degrees,
 abstract = {We study the role of caches in wireless interference networks. We focus on content caching and delivery across a Gaussian interference network, where both transmitters and receivers are equipped with caches. We provide a constant-factor approximation of the system's degrees of freedom (DoF), for arbitrary number of transmitters, number of receivers, content library size, receiver cache size, and transmitter cache size (as long as the transmitters combined can store the entire content library among them). We demonstrate approximate optimality with respect to information-theoretic bounds that do not impose any restrictions on the caching and delivery strategies. Our characterization reveals three key insights. First, the approximate DoF is achieved using a strategy that separates the physical and network layers. This separation architecture is thus approximately optimal. Second, we show that increasing transmitter cache memory beyond what is needed to exactly store the entire library between all transmitters does not provide more than a constant-factor benefit to the DoF. A consequence is that transmit zero-forcing is not needed for approximate optimality. Third, we derive an interesting tradeoff between the receiver memory and the number of transmitters needed for approximately maximal performance. In particular, if each receiver can store a constant fraction of the content library, then only a constant number of transmitters are needed. Our solution to the caching problem requires formulating and solving a new communication problem, the symmetric multiple multicast X-channel, for which we provide an exact DoF characterization.},
 author = {Hachem, Jad and Niesen, Urs and Diggavi, Suhas N},
 journal = {IEEE Transactions on Information Theory},
 number = {7},
 pages = {5359--5380},
 publisher = {IEEE},
 tags = {journal,CCWN,IT,WiNetnew,SCS,ANIT},
 title = {Degrees of freedom of cache-aided wireless interference networks},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1606.03175},
 doi = {10.1109/TIT.2018.2825321},
 volume = {64},
 year = {2018}
}

We study the role of caches in wireless interference networks. We focus on content caching and delivery across a Gaussian interference network, where both transmitters and receivers are equipped with caches. We provide a constant-factor approximation of the system's degrees of freedom (DoF), for arbitrary number of transmitters, number of receivers, content library size, receiver cache size, and transmitter cache size (as long as the transmitters combined can store the entire content library among them). We demonstrate approximate optimality with respect to information-theoretic bounds that do not impose any restrictions on the caching and delivery strategies. Our characterization reveals three key insights. First, the approximate DoF is achieved using a strategy that separates the physical and network layers. This separation architecture is thus approximately optimal. Second, we show that increasing transmitter cache memory beyond what is needed to exactly store the entire library between all transmitters does not provide more than a constant-factor benefit to the DoF. A consequence is that transmit zero-forcing is not needed for approximate optimality. Third, we derive an interesting tradeoff between the receiver memory and the number of transmitters needed for approximately maximal performance. In particular, if each receiver can store a constant fraction of the content library, then only a constant number of transmitters are needed. Our solution to the caching problem requires formulating and solving a new communication problem, the symmetric multiple multicast X-channel, for which we provide an exact DoF characterization.

@inproceedings{8443720,
 abstract = {We study a crowdsourcing aided road traffic estimation setup, where a fraction of users (vehicles) are malicious, and report wrong sensory information, or even worse, report the presence of Sybil (ghost) vehicles that do not physically exist. The motivation for such attacks lies in the possibility of creating a "virtual" congestion that can influence routing algorithms, leading to "actual" congestion and chaos. We propose a Sybil attack-resilient traffic estimation and routing algorithm that is resilient against such attacks. In particular, our algorithm leverages noisy information from legacy sensing infrastructure, along with the dynamics and proximity graph of vehicles inferred from crowdsourced data. Furthermore, the scalability of our algorithm is based on efficient Boolean Satisfiability (SAT) solvers. We validated our algorithm using real traffic data from the Italian city of Bologna. Our algorithm led to a significant reduction in average travel time in the presence of Sybil attacks, including cases where the travel time was reduced from about an hour to a few minutes.},
 author = {Y. {Shoukry} and S. {Mishra} and Z. {Luo} and S. {Diggavi}},
 booktitle = {2018 ACM/IEEE 9th International Conference on Cyber-Physical Systems (ICCPS)},
 doi = {10.1109/ICCPS.2018.00013},
 issn = {},
 keywords = {computability;graph theory;road traffic;telecommunication network routing;telecommunication security;crowdsourced data;traffic data;Sybil attack resilient traffic networks;physics-based trust propagation approach;crowdsourcing aided road traffic estimation setup;Sybil vehicles;virtual congestion;routing algorithms;chaos;routing algorithm;sensory information;noisy information;real traffic data;proximity graph;Automobiles;Routing;Roads;Heuristic algorithms;Estimation;Intelligent sensors;Secure Smart transportation systems;Sybil attacks;resilient routing},
 month = {April},
 number = {},
 pages = {43-54},
 tags = {conf,CPS},
 title = {Sybil Attack Resilient Traffic Networks: A Physics-Based Trust Propagation Approach},
 type = {4},
 volume = {},
 year = {2018}
}

We study a crowdsourcing aided road traffic estimation setup, where a fraction of users (vehicles) are malicious, and report wrong sensory information, or even worse, report the presence of Sybil (ghost) vehicles that do not physically exist. The motivation for such attacks lies in the possibility of creating a "virtual" congestion that can influence routing algorithms, leading to "actual" congestion and chaos. We propose a Sybil attack-resilient traffic estimation and routing algorithm that is resilient against such attacks. In particular, our algorithm leverages noisy information from legacy sensing infrastructure, along with the dynamics and proximity graph of vehicles inferred from crowdsourced data. Furthermore, the scalability of our algorithm is based on efficient Boolean Satisfiability (SAT) solvers. We validated our algorithm using real traffic data from the Italian city of Bologna. Our algorithm led to a significant reduction in average travel time in the presence of Sybil attacks, including cases where the travel time was reduced from about an hour to a few minutes.

@inproceedings{abdelzaher2018will,
 abstract = {An upcoming frontier for distributed computing might literally save lives in future military operations. In civilian scenarios, significant efficiencies were gained from interconnecting devices into networked services and applications that automate much of everyday life from smart homes to intelligent transportation. The ecosystem of such applications and services is collectively called the Internet of Things (IoT). Can similar benefits be gained in a military context by developing an IoT for the battlefield? This paper describes unique challenges in such a context as well as potential risks, mitigation strategies, and benefits.},
 author = {Abdelzaher, Tarek and Ayanian, Nora and Basar, Tamer and Diggavi, Suhas and Diesner, Jana and Ganesan, Deepak and Govindan, Ramesh and Jha, Susmit and Lepoint, Tancrede and Marlin, Ben and others},
 booktitle = {2018 IEEE 38th International Conference on Distributed Computing Systems (ICDCS)},
 organization = {IEEE},
 pages = {1129--1138},
 tags = {conf,CPS,DML},
 title = {Will distributed computing revolutionize peace? the emergence of battlefield iot},
 type = {4},
 doi = {10.1109/ICDCS.2018.00112},
 year = {2018}
}

An upcoming frontier for distributed computing might literally save lives in future military operations. In civilian scenarios, significant efficiencies were gained from interconnecting devices into networked services and applications that automate much of everyday life from smart homes to intelligent transportation. The ecosystem of such applications and services is collectively called the Internet of Things (IoT). Can similar benefits be gained in a military context by developing an IoT for the battlefield? This paper describes unique challenges in such a context as well as potential risks, mitigation strategies, and benefits.

@inproceedings{agarwal2018distorting,
 abstract = {In Cyber- Physical Systems (CPSs), inference based on communicated data is of critical significance as it can be used to manipulate or damage the control operations by adversaries. This calls for efficient mechanisms for secure transmission of data since control systems are becoming increasingly distributed over larger geographical areas. Distortion based security, recently proposed as one candidate for CPSs security, is not only more appropriate for these applications but also quite frugal in terms of prior requirements on shared keys. In this paper, we propose distortion-based metrics to protect CPSs communication and show that it is possible to confuse adversaries with just a few bits of pre-shared keys.},
 author = {Agarwal, Gaurav Kumar and Karmoose, Mohammed and Diggavi, Suhas and Fragouli, Christina and Tabuada, Paulo},
 booktitle = {2018 IEEE Conference on Decision and Control (CDC)},
 organization = {IEEE},
 pages = {1476--1481},
 tags = {conf,CPS},
 title = {Distorting an adversary's view in cyber-physical systems},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/1809.04580},
 doi = {10.1109/CDC.2018.8619457},
 year = {2018}
}

In Cyber- Physical Systems (CPSs), inference based on communicated data is of critical significance as it can be used to manipulate or damage the control operations by adversaries. This calls for efficient mechanisms for secure transmission of data since control systems are becoming increasingly distributed over larger geographical areas. Distortion based security, recently proposed as one candidate for CPSs security, is not only more appropriate for these applications but also quite frugal in terms of prior requirements on shared keys. In this paper, we propose distortion-based metrics to protect CPSs communication and show that it is possible to confuse adversaries with just a few bits of pre-shared keys.

@inproceedings{data2018data,
 abstract = {We consider distributed gradient computation, where both data and computation are distributed among m worker machines, t of which can be Byzantine adversaries, and a designated (master) node computes the model/parameter vector, iteratively using gradient descent (GD). The Byzantine adversaries can (collaboratively) deviate arbitrarily from their gradient computation. To solve this, we propose a method based on data encoding and (real) error correction to combat the adversarial behavior. We can tolerate up to t ≤ [m-1/2] corrupt worker nodes, which is information-theoretically optimal. Our method does not assume any probability distribution on the data. We develop a sparse encoding scheme which enables computationally efficient data encoding. We demonstrate a trade-off between the number of adversaries tolerated and the resource requirement (storage and computational complexity). As an example, our scheme incurs a constant overhead (storage and computational complexity) over that required by the distributed GD algorithm, without adversaries, for t ≤ m/3.},
 author = {Data, Deepesh and Song, Linqi and Diggavi, Suhas},
 booktitle = {2018 56th Annual Allerton Conference on Communication, Control, and Computing (Allerton)},
 organization = {IEEE},
 pages = {863--870},
 tags = {conf,SDL,DML},
 title = {Data encoding for Byzantine-resilient distributed gradient descent},
 type = {4},
 doi = {10.1109/ALLERTON.2018.8636017},
 year = {2018}
}

We consider distributed gradient computation, where both data and computation are distributed among m worker machines, t of which can be Byzantine adversaries, and a designated (master) node computes the model/parameter vector, iteratively using gradient descent (GD). The Byzantine adversaries can (collaboratively) deviate arbitrarily from their gradient computation. To solve this, we propose a method based on data encoding and (real) error correction to combat the adversarial behavior. We can tolerate up to t ≤ [m-1/2] corrupt worker nodes, which is information-theoretically optimal. Our method does not assume any probability distribution on the data. We develop a sparse encoding scheme which enables computationally efficient data encoding. We demonstrate a trade-off between the number of adversaries tolerated and the resource requirement (storage and computational complexity). As an example, our scheme incurs a constant overhead (storage and computational complexity) over that required by the distributed GD algorithm, without adversaries, for t ≤ m/3.

@inproceedings{showkatbakhsh2018privacy,
 abstract = {Data privacy is an important concern in machine learning, and is fundamentally at odds with the task of training useful learning models, which typically require acquisition of large amounts of private user data. One possible way of fulfilling the machine learning task while preserving user privacy is to train the model on a transformed, noisy version of the data, which does not reveal the data itself directly to the training procedure. In this work, we analyze the privacy-utility tradeoff of two such schemes for the problem of linear regression: additive noise, and random projections. In contrast to previous work, we consider a recently proposed notion of differential privacy that is based on conditional mutual information (MI-DP), which is stronger than the conventional (ε,δ) -differential privacy, and use relative objective error as the utility metric. We find that projecting the data to a lower-dimensional subspace before adding noise attains a better trade-off in general. We also make a connection between privacy problem and (non-coherent) SIMO, which has been extensively studied in wireless communication, and use tools from there for the analysis. We present numerical results demonstrating the performance of the schemes.},
 author = {Showkatbakhsh, Mehrdad and Karakus, Can and Diggavi, Suhas},
 booktitle = {2018 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {186--190},
 tags = {conf,PDL},
 title = {Privacy-Utility Trade-off of Linear Regression under Random Projections and Additive Noise},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/1902.04688},
 doi = {10.1109/ISIT.2018.8437722},
 year = {2018}
}

Data privacy is an important concern in machine learning, and is fundamentally at odds with the task of training useful learning models, which typically require acquisition of large amounts of private user data. One possible way of fulfilling the machine learning task while preserving user privacy is to train the model on a transformed, noisy version of the data, which does not reveal the data itself directly to the training procedure. In this work, we analyze the privacy-utility tradeoff of two such schemes for the problem of linear regression: additive noise, and random projections. In contrast to previous work, we consider a recently proposed notion of differential privacy that is based on conditional mutual information (MI-DP), which is stronger than the conventional (ε,δ) -differential privacy, and use relative objective error as the utility metric. We find that projecting the data to a lower-dimensional subspace before adding noise attains a better trade-off in general. We also make a connection between privacy problem and (non-coherent) SIMO, which has been extensively studied in wireless communication, and use tools from there for the analysis. We present numerical results demonstrating the performance of the schemes.

@inproceedings{srinivasavaradhan2018maximum,
 abstract = {The problem of reconstructing a sequence when observed through multiple looks over deletion channels occurs in “de novo” DNA sequencing. The DNA could be sequenced multiple times, yielding several “looks” of it, but each time the sequencer could be noisy with (independent) deletion impairments. The main goal of this paper is to develop reconstruction algorithms for a sequence observed through the lens of a fixed number of deletion channels. We use the probabilistic model of the deletion channels to develop both symbol-wise and sequence maximum likelihood decoding criteria, and algorithms motivated by them. Numerical evaluations demonstrate improvement in terms of edit distance error, over earlier algorithms.},
 author = {Srinivasavaradhan, Sundara Rajan and Du, Michelle and Diggavi, Suhas and Fragouli, Christina},
 booktitle = {2018 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {436--440},
 tags = {conf,BioInf,IT,NDS},
 title = {On maximum likelihood reconstruction over multiple deletion channels},
 type = {4},
 doi = {10.1109/ISIT.2018.8437519},
 year = {2018}
}

The problem of reconstructing a sequence when observed through multiple looks over deletion channels occurs in “de novo” DNA sequencing. The DNA could be sequenced multiple times, yielding several “looks” of it, but each time the sequencer could be noisy with (independent) deletion impairments. The main goal of this paper is to develop reconstruction algorithms for a sequence observed through the lens of a fixed number of deletion channels. We use the probabilistic model of the deletion channels to develop both symbol-wise and sequence maximum likelihood decoding criteria, and algorithms motivated by them. Numerical evaluations demonstrate improvement in terms of edit distance error, over earlier algorithms.

@inproceedings{sultangazin2018protecting,
 abstract = {The vision of an Internet-of-Things calls for combining the increasing connectivity of devices at the edge with the ability to compute either at the edge or on more powerful servers in the network. There is great interest in exploring the feasibility of these ideas when devices such as quadcopters or ground robots at the edge are controlled over the cloud, i.e., by leveraging computational power available elsewhere in the network. One of the main difficulties, especially in the context of the Internet-of-Battlefield- Things is the need to keep the data private. In this paper we propose a solution to this problem by extending previous results by the authors from a single system controlled over the cloud to networks of systems that are controlled and coordinated over the cloud. We propose a noncryptographic lightweight encoding scheme that ensures the privacy of the data exchanged by all the participating parties.},
 author = {Sultangazin, Alimzhan and Diggavi, Suhas and Tabuada, Paulo},
 booktitle = {2018 27th International Conference on Computer Communication and Networks (ICCCN)},
 organization = {IEEE},
 pages = {1--7},
 tags = {conf,CPS},
 title = {Protecting the privacy of networked multi-agent systems controlled over the cloud},
 type = {4},
 doi = {10.1109/ICCCN.2018.8487355},
 year = {2018}
}

The vision of an Internet-of-Things calls for combining the increasing connectivity of devices at the edge with the ability to compute either at the edge or on more powerful servers in the network. There is great interest in exploring the feasibility of these ideas when devices such as quadcopters or ground robots at the edge are controlled over the cloud, i.e., by leveraging computational power available elsewhere in the network. One of the main difficulties, especially in the context of the Internet-of-Battlefield- Things is the need to keep the data private. In this paper we propose a solution to this problem by extending previous results by the authors from a single system controlled over the cloud to networks of systems that are controlled and coordinated over the cloud. We propose a noncryptographic lightweight encoding scheme that ensures the privacy of the data exchanged by all the participating parties.

@article{Joyson_twc17,
 author = {Sebastian, J. and Karakus, C. and Diggavi, S. N.},
 file = {:papers:twc17_ic_p2p.pdf},
 journal = {Preprint},
 month = {June},
 tags = {journalSub,WiIntMgmt,IT,WiNet},
 title = {Approximate Capacity of Fading Gaussian Interference Channels with Point-to-Point Codes},
 type = {1},
 year = {2017}
}

@article{7865913,
 author = {J. Hachem and N. Karamchandani and S. N. Diggavi},
 doi = {10.1109/TIT.2017.2664817},
 file = {:papers:jad_multilevel.pdf},
 issn = {0018-9448},
 journal = {IEEE Transactions on Information Theory},
 keywords = {cache storage;cellular radio;content management;cache memory sharing;coded caching-and-delivery scheme;joint content delivery and storage design;local-coverage wireless access point deployment;multilevel access;multilevel popularity;non-cut-set bound;order-optimal scheme;sliding-window entropy inequality;sparsely distributed large-coverage macrocell base station;wireless architecture;wireless content;Cache memory;Computer architecture;Context;IEEE 802.11 Standard;Internet;Wireless networks;Content distribution networks;cache memory;network coding},
 month = {May},
 number = {5},
 pages = {3108-3141},
 tags = {journal,IT,Caching,WiNet},
 title = {Coded Caching for Multi-level Popularity and Access},
 type = {2},
 volume = {63},
 year = {2017}
}

@article{7911284,
 abstract = {We study parallel two-user interference channels when the interference is bursty and feedback is available from the respective receivers. Presence of interference in each subcarrier is modeled as a memoryless Bernoulli random state. The states across subcarriers are drawn from an arbitrary joint distribution with the same marginal probability for each subcarrier and instantiated independent and identically distributed (i.i.d.) over time. For the linear deterministic setup with symmetric interference in each subcarrier, we give a complete characterization of the capacity region. For the analogous setup with Gaussian noise, we give outer bounds and a tight generalized degrees of freedom characterization. We propose a novel helping mechanism, which enables subcarriers in very strong interference regime to help in recovering interfered signals for subcarriers in strong and weak interference regimes. Depending on the interference and burstiness regime, the inner bounds either employ the proposed helping mechanism to code across subcarriers or treat the subcarriers separately. The outer bounds demonstrate a connection to a subset entropy inequality by Madiman and Tetali.},
 author = {S. {Mishra} and I. {Wang} and S. N. {Diggavi}},
 doi = {10.1109/TIT.2017.2697866},
 issn = {1557-9654},
 journal = {IEEE Transactions on Information Theory},
 keywords = {Gaussian noise;probability;radio networks;radiofrequency interference;harnessing bursty interference;multicarrier systems;output feedback;interference channels;respective receivers;memoryless Bernoulli random state;arbitrary joint distribution;probability;symmetric interference;capacity region;Gaussian noise;subset entropy inequality;wireless networks;Receivers;Interference channels;Output feedback;Encoding;Channel models;Niobium;Interference channel;bursty interference;feedback;multicarrier systems},
 month = {July},
 number = {7},
 pages = {4430-4452},
 tags = {journal,IT,ANIT,WNIF},
 title = {Harnessing Bursty Interference in Multicarrier Systems With Output Feedback},
 type = {2},
 volume = {63},
 year = {2017}
}

We study parallel two-user interference channels when the interference is bursty and feedback is available from the respective receivers. Presence of interference in each subcarrier is modeled as a memoryless Bernoulli random state. The states across subcarriers are drawn from an arbitrary joint distribution with the same marginal probability for each subcarrier and instantiated independent and identically distributed (i.i.d.) over time. For the linear deterministic setup with symmetric interference in each subcarrier, we give a complete characterization of the capacity region. For the analogous setup with Gaussian noise, we give outer bounds and a tight generalized degrees of freedom characterization. We propose a novel helping mechanism, which enables subcarriers in very strong interference regime to help in recovering interfered signals for subcarriers in strong and weak interference regimes. Depending on the interference and burstiness regime, the inner bounds either employ the proposed helping mechanism to code across subcarriers or treat the subcarriers separately. The outer bounds demonstrate a connection to a subset entropy inequality by Madiman and Tetali.

@article{7920290,
 abstract = {We investigate whether uncoded schemes are optimal for Gaussian sources on multiuser Gaussian channels. Particularly, we consider two problems: the first is to send correlated Gaussian sources on a Gaussian broadcast channel where each receiver is interested in reconstructing only one source component (or one specific linear function of the sources) under the mean squared error distortion measure; the second is to send correlated Gaussian sources on a Gaussian multiple-access channel, where each transmitter observes a noisy combination of the sources, and the receiver wishes to reconstruct the individual source components (or individual linear functions) under the mean squared error distortion measure. It is shown that when the channel parameters satisfy certain general conditions, the induced distortion tuples are on the boundary of the achievable distortion region, and thus optimal. Instead of following the conventional approach of attempting to characterize the achievable distortion region, we ask the question whether and how a match can be effectively determined. This decision problem formulation helps to circumvent the difficult optimization problem often embedded in region characterization problems, and it also leads us to focus on the critical conditions in the outer bounds that make the inequalities become equalities, which effectively decouple the overall problem into several simpler sub-problems. Optimality results previously unknown in the literature are obtained using this novel approach. Explicit and novel outer bounds are derived for the two problems as the byproducts of our investigation.},
 author = {C. {Tian} and J. {Chen} and S. N. {Diggavi} and S. {Shamai Shitz}},
 doi = {10.1109/TIT.2017.2701807},
 issn = {1557-9654},
 journal = {IEEE Transactions on Information Theory},
 keywords = {broadcast channels;Gaussian channels;mean square error methods;multiuser channels;matched multiuser Gaussian source channel communications;uncoded schemes;Gaussian broadcast channel;mean squared error distortion measure;correlated Gaussian sources;Gaussian multiple-access channel;mean squared error distortion;distortion tuples;optimization problem;region characterization problems;Distortion;Distortion measurement;Optimization;Receivers;Noise measurement;Decoding;Encoding;Broadcast;multiple access;combined source-channel coding},
 month = {July},
 number = {7},
 pages = {4155-4171},
 tags = {journal,IT,SCS},
 title = {Matched Multiuser Gaussian Source Channel Communications via Uncoded Schemes},
 type = {2},
 volume = {63},
 year = {2017}
}

We investigate whether uncoded schemes are optimal for Gaussian sources on multiuser Gaussian channels. Particularly, we consider two problems: the first is to send correlated Gaussian sources on a Gaussian broadcast channel where each receiver is interested in reconstructing only one source component (or one specific linear function of the sources) under the mean squared error distortion measure; the second is to send correlated Gaussian sources on a Gaussian multiple-access channel, where each transmitter observes a noisy combination of the sources, and the receiver wishes to reconstruct the individual source components (or individual linear functions) under the mean squared error distortion measure. It is shown that when the channel parameters satisfy certain general conditions, the induced distortion tuples are on the boundary of the achievable distortion region, and thus optimal. Instead of following the conventional approach of attempting to characterize the achievable distortion region, we ask the question whether and how a match can be effectively determined. This decision problem formulation helps to circumvent the difficult optimization problem often embedded in region characterization problems, and it also leads us to focus on the critical conditions in the outer bounds that make the inequalities become equalities, which effectively decouple the overall problem into several simpler sub-problems. Optimality results previously unknown in the literature are obtained using this novel approach. Explicit and novel outer bounds are derived for the two problems as the byproducts of our investigation.

@article{7942013,
 abstract = {We propose a cellular architecture that combines multiuser MIMO downlink with opportunistic use of unlicensed Industrial, Scientific, and Medical Radio (ISM) bands to establish device-to-device (D2D) cooperation. The architecture consists of a physical-layer cooperation scheme based on forming downlink virtual MIMO channels through D2D relaying, and a novel resource allocation strategy for such D2D-enabled networks. We prove the approximate optimality of the physical-layer scheme, and demonstrate that such cooperation boosts the effective SNR of the weakest user in the system, especially in the many-user regime, due to multiuser diversity. To harness this physical-layer scheme, we formulate the cooperative user scheduling and the relay selection problem using the network utility maximization framework. For such a cooperative network, we propose a novel utility metric that jointly captures fairness in throughput and the cost of relaying in the system. We propose a joint user scheduling and relay selection algorithm, which we prove to be asymptotically optimal. We study the architecture through system-level simulations over a wide range of scenarios. The highlight of these simulations is an approximately 6× improvement in data rate for cell-edge (bottom fifth-percentile) users (over the state-of-the-art) while still improving the overall throughput, and considering various system constraints.},
 author = {C. {Karakus} and S. {Diggavi}},
 doi = {10.1109/TWC.2017.2712649},
 issn = {1558-2248},
 journal = {IEEE Transactions on Wireless Communications},
 keywords = {cellular radio;cooperative communication;MIMO communication;multiuser channels;relay networks (telecommunication);resource allocation;telecommunication scheduling;cellular architecture;multiuser MIMO downlink;unlicensed Industrial Scientific and Medical Radio bands;unlicensed ISM bands;device-to-device cooperation;D2D cooperation;physical-layer cooperation scheme;downlink virtual MIMO channels;D2D relaying;resource allocation strategy;D2D-enabled networks;physical-layer scheme;many-user regime;multiuser diversity;cooperative user scheduling;network utility maximization framework;joint user scheduling algorithm;relay selection algorithm;system-level simulations;cell-edge users;Device-to-device communication;Downlink;Computer architecture;Throughput;MIMO;Relays;Base stations;D2D;opportunistic scheduling;multiuser MIMO;ISM bands;user cooperation},
 month = {Sep.},
 number = {9},
 pages = {5616-5629},
 tags = {journal,WiNetnew,WiNet},
 title = {Enhancing Multiuser MIMO Through Opportunistic D2D Cooperation},
 type = {2},
 volume = {16},
 year = {2017}
}

We propose a cellular architecture that combines multiuser MIMO downlink with opportunistic use of unlicensed Industrial, Scientific, and Medical Radio (ISM) bands to establish device-to-device (D2D) cooperation. The architecture consists of a physical-layer cooperation scheme based on forming downlink virtual MIMO channels through D2D relaying, and a novel resource allocation strategy for such D2D-enabled networks. We prove the approximate optimality of the physical-layer scheme, and demonstrate that such cooperation boosts the effective SNR of the weakest user in the system, especially in the many-user regime, due to multiuser diversity. To harness this physical-layer scheme, we formulate the cooperative user scheduling and the relay selection problem using the network utility maximization framework. For such a cooperative network, we propose a novel utility metric that jointly captures fairness in throughput and the cost of relaying in the system. We propose a joint user scheduling and relay selection algorithm, which we prove to be asymptotically optimal. We study the architecture through system-level simulations over a wide range of scenarios. The highlight of these simulations is an approximately 6× improvement in data rate for cell-edge (bottom fifth-percentile) users (over the state-of-the-art) while still improving the overall throughput, and considering various system constraints.

@article{mishra2017secure,
 author = {Mishra, Shaunak and Shoukry, Yasser and Karamchandani, Nikhil and Diggavi, Suhas N and Tabuada, Paulo},
 file = {:papers:mishra_secure.pdf},
 journal = {IEEE Transactions on Control of Network Systems},
 number = {1},
 pages = {49--59},
 publisher = {IEEE},
 tags = {journal,IT,SecureCPS},
 title = {Secure state estimation against sensor attacks in the presence of noise},
 type = {2},
 volume = {4},
 year = {2017}
}

@article{mishra2017wiretapped,
 author = {Mishra, Manoj and Dey, Bikash Kumar and Prabhakaran, Vinod M and Diggavi, Suhas N},
 journal = {IEEE Transactions on Information Theory},
 number = {4},
 pages = {2560--2595},
 publisher = {IEEE},
 tags = {journal,IT,WNS},
 title = {Wiretapped Oblivious Transfer},
 type = {2},
 volume = {63},
 year = {2017}
}

@article{siavoshani2017multi,
 author = {Siavoshani, Mahdi Jafari and Mishra, Shaunak and Fragouli, Christina and Diggavi, Suhas N},
 file = {:papers:siavoshani_secretkey.pdf},
 journal = {IEEE Transactions on Information Forensics and Security},
 number = {2},
 pages = {323--337},
 publisher = {IEEE},
 tags = {journal,IT,WiNetSec},
 title = {Multi-Party Secret Key Agreement Over State-Dependent Wireless Broadcast Channels},
 type = {2},
 volume = {12},
 year = {2017}
}

@inproceedings{adaptive,
 author = {Jad Hachem and Nikhil Karamchandani and Sharayu Moharir and Suhas Diggavi},
 booktitle = {2017 IEEE International Symposium on Information Theory (ISIT)},
 file = {:papers:jad_adaptive.pdf},
 month = {June},
 note = {available on arXiv:1705.03852 [cs.IT]},
 tags = {conf,IT,Caching,WiNet},
 title = {Coded Caching With Partial Adaptive Matching},
 type = {4},
 year = {2017}
}

@inproceedings{dist_opt_isit,
 author = {Karakus, C.  and Sun, Y. and Diggavi, S. N.},
 booktitle = {2017 IEEE International Symposium on Information Theory},
 file = {:papers:isit17_dist_opt.pdf},
 month = {June},
 tags = {conf,IT,optimization},
 title = {Encoded Distributed Optimization},
 type = {4},
 year = {2017}
}

@inproceedings{hachem2017caching,
 author = {Hachem, Jad and Karamchandani, Nikhil and Moharir, Sharayu and Diggavi, Suhas},
 booktitle = {2017 IEEE Information Theory Workshop (ITW)},
 organization = {IEEE},
 pages = {61--65},
 tags = {conf,CCWN,IT},
 title = {Caching with partial matching under Zipf demands},
 type = {4},
 year = {2017}
}

@inproceedings{Joyson_isit17,
 author = {Sebastian, J. and Sengupta, A. and Diggavi, S. N.},
 booktitle = {2017 IEEE International Symposium on Information Theory (ISIT)},
 file = {:papers:isit17_noncoherent.pdf},
 month = {June},
 tags = {conf,WiIntMgmt,IT,WiNet,Noncoherent},
 title = {On Capacity of Noncoherent MIMO with Asymmetric Link Strengths},
 type = {4},
 year = {2017}
}

@inproceedings{karakus2017straggler,
 author = {Karakus, Can and Sun, Yifan and Diggavi, Suhas and Yin, Wotao},
 booktitle = {Advances in Neural Information Processing Systems},
 pages = {5434--5442},
 tags = {conf,DML,EncOpt},
 title = {Straggler mitigation in distributed optimization through data encoding},
 type = {4},
 year = {2017}
}

@conference{Mao-Diggavi-Kannan-nanopore-isit,
 author = {Mao, Wei and Diggavi, Suhas and Kannan, Sreeram},
 booktitle = {2017 IEEE International Symposium on Information Theory Proceedings (ISIT)},
 file = {:papers:isit_nanopore.pdf},
 tags = {conf,IT,Nanopore},
 title = {Models and information-theoretic bounds for nanopore sequencing},
 type = {4},
 year = {2017}
}

@inproceedings{PCISIT2017,
 author = {Y. Yona and S. Diggavi},
 booktitle = {2017 IEEE International Symposium on Information Theory Proceedings},
 file = {:papers:pcisit2017.pdf},
 tags = {conf,IT,Hardware},
 title = {The Effect of Bias on the Guesswork of Hash Functions},
 type = {4},
 year = {2017}
}

@inproceedings{showkatbakhsh2017smt,
 author = {Showkatbakhsh, Mehrdad and Shoukry, Yasser and Chen, Robert H and Diggavi, Suhas and Tabuada, Paulo},
 booktitle = {2017 IEEE 56th Annual Conference on Decision and Control (CDC)},
 organization = {IEEE},
 pages = {157--162},
 tags = {conf,CPS},
 title = {An SMT-based approach to secure state estimation under sensor and actuator attacks},
 type = {4},
 year = {2017}
}

@inproceedings{tsai2017distortion,
 author = {Tsai, Chi-Yo and Agarwal, Gaurav Kumar and Fragouli, Christina and Diggavi, Suhas},
 booktitle = {2017 IEEE international symposium on information theory (ISIT)},
 organization = {IEEE},
 pages = {1913--1917},
 tags = {conf,CPS},
 title = {A distortion based approach for protecting inferences},
 type = {4},
 year = {2017}
}

@inproceedings{wang2017implementation,
 author = {Wang, Wei-Che and Yona, Yair and Wu, Yizhang and Hung, Szu-Yao and Diggavi, Suhas and Gupta, Puneet},
 booktitle = {2017 Asian Hardware Oriented Security and Trust Symposium (AsianHOST)},
 organization = {IEEE},
 pages = {13--18},
 tags = {conf,HS},
 title = {Implementation of stable PUFs using gate oxide breakdown},
 type = {4},
 year = {2017}
}

@article{PasswordCrackin2016,
 author = {Yona, Yair and Diggavi, Suhas},
 file = {:papers:passwordcrackin2016.pdf},
 journal = {Preprint},
 tags = {journalSub,IT,Hardware},
 title = {Password Cracking: The Effect of Bias on the Average Guesswork of Hash Functions},
 type = {1},
 year = {2016}
}

@article{Phisical_Challange_Response_Auth2016,
 abstract = {Embedded sensing systems are pervasively used in life- and security-critical systems such as those found in airplanes, automobiles, and healthcare. Traditional security mechanisms for these sensors focus on data encryption and other post-processing techniques, but the sensors themselves often remain vulnerable to attacks in the physical/analog domain. If an adversary manipulates a physical/analog signal prior to digitization, no amount of digital security mechanisms after the fact can help. Fortunately, nature imposes fundamental constraints on how these analog signals can behave. This work presents PyCRA, a physical challenge-response authentication scheme designed to protect active sensing systems against physical attacks occurring in the analog domain. PyCRA provides security for active sensors by continually challenging the surrounding environment via random but deliberate physical probes. By analyzing the responses to these probes, and by using the fact that the adversary cannot change the underlying laws of physics, we provide an authentication mechanism that not only detects malicious attacks but provides resilience against them. We demonstrate the effectiveness of PyCRA through several case studies using two sensing systems: (1) magnetic sensors like those found wheel speed sensors in robotics and automotive, and (2) commercial RFID tags used in many security-critical applications. Finally, we outline methods and theoretical proofs for further enhancing the resilience of PyCRA to active attacks by means of a confusion phase---a period of low signal to noise ratio that makes it more difficult for an attacker to correctly identify and respond to PyCRA's physical challenges. In doing so, we evaluate both the robustness and the limitations of PyCRA, concluding by outlining practical considerations as well as further applications for the proposed authentication mechanism.},
 author = {Shoukry, Yasser and Martin, Paul and Yona, Yair and Diggavi, Suhas and Srivastava, Mani},
 file = {:papers:physical_challange_response_auth2016.pdf},
 journal = {Preprint},
 tags = {journalSub,IT,Hardware},
 title = {Attack Resilience and Recovery using Physical Challenge Response Authentication for Active Sensors Under Integrity Attacks},
 type = {1},
 url_arxiv = {https://arxiv.org/abs/1605.02062},
 year = {2016}
}

Embedded sensing systems are pervasively used in life- and security-critical systems such as those found in airplanes, automobiles, and healthcare. Traditional security mechanisms for these sensors focus on data encryption and other post-processing techniques, but the sensors themselves often remain vulnerable to attacks in the physical/analog domain. If an adversary manipulates a physical/analog signal prior to digitization, no amount of digital security mechanisms after the fact can help. Fortunately, nature imposes fundamental constraints on how these analog signals can behave. This work presents PyCRA, a physical challenge-response authentication scheme designed to protect active sensing systems against physical attacks occurring in the analog domain. PyCRA provides security for active sensors by continually challenging the surrounding environment via random but deliberate physical probes. By analyzing the responses to these probes, and by using the fact that the adversary cannot change the underlying laws of physics, we provide an authentication mechanism that not only detects malicious attacks but provides resilience against them. We demonstrate the effectiveness of PyCRA through several case studies using two sensing systems: (1) magnetic sensors like those found wheel speed sensors in robotics and automotive, and (2) commercial RFID tags used in many security-critical applications. Finally, we outline methods and theoretical proofs for further enhancing the resilience of PyCRA to active attacks by means of a confusion phase—a period of low signal to noise ratio that makes it more difficult for an attacker to correctly identify and respond to PyCRA's physical challenges. In doing so, we evaluate both the robustness and the limitations of PyCRA, concluding by outlining practical considerations as well as further applications for the proposed authentication mechanism.

@article{bidokhti2016capacity,
 author = {Bidokhti, Shirin Saeedi and Prabhakaran, Vinod M and Diggavi, Suhas N},
 journal = {IEEE Transactions on Information Theory},
 number = {9},
 pages = {4968--4992},
 publisher = {IEEE},
 tags = {journal,IT},
 title = {Capacity results for multicasting nested message sets over combination networks},
 type = {2},
 volume = {62},
 year = {2016}
}

@article{czap2016lp,
 author = {Czap, L{\'a}szl{\'o} and Prabhakaran, Vinod M and Fragouli, Christina and Diggavi, Suhas N},
 journal = {IEEE Transactions on Information Theory},
 number = {5},
 pages = {2430--2480},
 publisher = {IEEE},
 tags = {journal,IT,WNS},
 title = {An LP Characterization of the Secret-message Capacity of Three Erasure Networks With Feedback},
 type = {2},
 volume = {62},
 year = {2016}
}

@article{karamchandani2016hierarchical,
 author = {Karamchandani, Nikhil and Niesen, Urs and Maddah-Ali, Mohammad Ali and Diggavi, Suhas N},
 file = {:papers:karam_hier.pdf},
 journal = {IEEE Transactions on Information Theory},
 number = {6},
 pages = {3212--3229},
 publisher = {IEEE},
 tags = {journal,IT,Caching,WiNet},
 title = {Hierarchical coded caching},
 type = {2},
 volume = {62},
 year = {2016}
}

@inproceedings{7541332,
 author = {J. Hachem and U. Niesen and S. Diggavi},
 booktitle = {2016 IEEE International Symposium on Information Theory (ISIT)},
 doi = {10.1109/ISIT.2016.7541332},
 file = {:papers:jad_layered.pdf},
 keywords = {Gaussian channels;radio receivers;radio transmitters;radiofrequency interference;wireless channels;degrees-of-freedom;layered architecture;layered caching architecture;network layer;physical layer;radio receiver;radio transmitter;two-user Gaussian interference channel;Cache memory;Computer architecture;Interference channels;Physical layer;Receivers;Servers;Transmitters},
 month = {July},
 pages = {415-419},
 tags = {conf,IT,Caching,WiNet},
 title = {A layered caching architecture for the interference channel},
 type = {4},
 year = {2016}
}

@inproceedings{DBLP:conf/allerton/ShowkatbakhshTD16,
 abstract = {This work is concerned with the identification of linear time-invariant systems in the presence of an adversarial agent that attacks sensor measurements. The attacker is omniscient and we impose no restrictions (statistical or otherwise) on how the adversary alters the sensor measurements. We work in a noisy scenario where, in addition to the attacks, the sensor measurements are also affected by additive noise. Given a bound on the number of attacked sensors, and under a certain observability condition, we show that we can still construct a model that is useful for stabilization. Furthermore, we show that this model is closely related to the original system.},
 author = {Mehrdad Showkatbakhsh and
Paulo Tabuada and
Suhas N. Diggavi},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/conf/allerton/ShowkatbakhshTD16},
 booktitle = {54th Annual Allerton Conference on Communication, Control, and Computing,
Allerton 2016, Monticello, IL, USA, September 27-30, 2016},
 crossref = {DBLP:conf/allerton/2016},
 doi = {10.1109/ALLERTON.2016.7852363},
 file = {:papers:mehrdad_allerton16.pdf},
 pages = {1137--1141},
 tags = {conf,IT,SecureCPS},
 timestamp = {Fri, 17 Feb 2017 10:49:47 +0100},
 title = {Secure system identification},
 type = {4},
 year = {2016}
}

This work is concerned with the identification of linear time-invariant systems in the presence of an adversarial agent that attacks sensor measurements. The attacker is omniscient and we impose no restrictions (statistical or otherwise) on how the adversary alters the sensor measurements. We work in a noisy scenario where, in addition to the attacks, the sensor measurements are also affected by additive noise. Given a bound on the number of attacked sensors, and under a certain observability condition, we show that we can still construct a model that is useful for stabilization. Furthermore, we show that this model is closely related to the original system.

@inproceedings{DBLP:conf/cdc/ShowkatbakhshTD16,
 author = {Mehrdad Showkatbakhsh and
Paulo Tabuada and
Suhas N. Diggavi},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/conf/cdc/ShowkatbakhshTD16},
 booktitle = {55th {IEEE} Conference on Decision and Control, CDC 2016, Las Vegas,
NV, USA, December 12-14, 2016},
 crossref = {DBLP:conf/cdc/2016},
 doi = {10.1109/CDC.2016.7799376},
 file = {:papers:mehrdad_cdc16.pdf},
 pages = {7177--7182},
 tags = {conf,IT,SecureCPS},
 timestamp = {Tue, 03 Jan 2017 15:16:45 +0100},
 title = {System identification in the presence of adversarial outputs},
 type = {4},
 year = {2016}
}

@inproceedings{HOST2016,
 author = {W. C. Wang and Y. Yona and S. Diggavi and P. Gupta},
 booktitle = {2016 IEEE International Symposium on Hardware Oriented Security and Trust (HOST)},
 doi = {10.1109/HST.2016.7495551},
 file = {:papers:ledpuf.pdf},
 keywords = {Entropy;Hardware;Noise measurement;Semiconductor device measurement;Shape;Silicon;Temperature measurement},
 month = {May},
 pages = {25-30},
 tags = {conf,IT,Hardware},
 title = {LEDPUF: Stability-guaranteed physical unclonable functions through locally enhanced defectivity},
 type = {4},
 year = {2016}
}

@inproceedings{Joyson_AF_isit,
 author = {Sebastian, J. and Karakus, C. and Diggavi, S. N.},
 booktitle = {2016 IEEE International Symposium on Information Theory (ISIT)},
 file = {:papers:isit16_ic_p2p.pdf},
 month = {July},
 pages = {715-719},
 tags = {conf,WiIntMgmt,IT,WiNet,Noncoherent},
 title = {Approximately achieving the feedback interference channel capacity with point-to-point codes},
 type = {4},
 year = {2016}
}

@inproceedings{karamchandani2016rate,
 author = {Karamchandani, Nikhil and Diggavi, Suhas and Caire, Giuseppe and Shamai, Shlomo},
 booktitle = {2016 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {2724--2728},
 tags = {conf,CCWN},
 title = {Rate and delay for coded caching with carrier aggregation},
 type = {4},
 year = {2016}
}

@article{czap2015secure,
 author = {Czap, L{\'a}szl{\'o} and Fragouli, Christina and Prabhakaran, Vinod M and Diggavi, Suhas},
 journal = {IEEE Transactions on Information Theory},
 number = {4},
 pages = {1667--1686},
 publisher = {IEEE},
 tags = {journal,IT,WNS},
 title = {Secure network coding with erasures and feedback},
 type = {2},
 doi = {http://dx.doi.org/10.1109/TIT.2015.2402272},
 volume = {61},
 year = {2015}
}

@article{DBLP:journals/pieee/FragouliPCD15,
 author = {Christina Fragouli and
Vinod M. Prabhakaran and
L{\'{a}}szl{\'{o}} Czap and
Suhas N. Diggavi},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/journals/pieee/FragouliPCD15},
 file = {:papers:winetsec_boe.pdf},
 journal = {Proceedings of the IEEE},
 number = {10},
 pages = {1826--1840},
 tags = {journal,WiNetSec,IT},
 timestamp = {Tue, 22 Sep 2015 10:37:29 +0200},
 title = {Wireless Network Security: Building on Erasures},
 type = {2},
 doi = {http://dx.doi.org/10.1109/JPROC.2015.2438312},
 volume = {103},
 year = {2015}
}

@article{DBLP:journals/tit/CzapPFD15,
 abstract = {We consider a 1-to-K communication scenario, where a source transmits private messages to K receivers through a broadcast erasure channel, and the receivers feed back strictly causally and publicly their channel states after each transmission. We explore the achievable rate region when we require that the message to each receiver remains secret - in the information theoretical sense - from all the other receivers. We characterize the capacity of secure communication in all the cases where the capacity of the 1-to-K communication scenario without the requirement of security is known. As a special case, we characterize the secret-message capacity of a single receiver point-to-point erasure channel with public state-feedback in the presence of a passive eavesdropper. 
We find that in all cases where we have an exact characterization, we can achieve the capacity by using linear complexity two-phase schemes: in the first phase we create appropriate secret keys, and in the second phase we use them to encrypt each message. We find that the amount of key we need is smaller than the size of the message, and equal to the amount of encrypted message the potential eavesdroppers jointly collect. Moreover, we prove that a dishonest receiver that provides deceptive feedback cannot diminish the rate experienced by the honest receivers. 
We also develop a converse proof which reflects the two-phase structure of our achievability scheme. As a side result, our technique leads to a new outer bound proof for the non-secure communication problem.},
 author = {L{\'{a}}szl{\'{o}} Czap and
Vinod M. Prabhakaran and
Christina Fragouli and
Suhas N. Diggavi},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/journals/tit/CzapPFD15},
 doi = {10.1109/TIT.2015.2449838},
 file = {:papers:SecretCom_Czap.pdf},
 journal = {IEEE Transactions on Information Theory},
 number = {9},
 pages = {4788--4808},
 tags = {journal,WiNetSec,IT},
 timestamp = {Thu, 20 Aug 2015 14:54:38 +0200},
 title = {Secret Communication Over Broadcast Erasure Channels With State-Feedback},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1408.1800},
 volume = {61},
 year = {2015}
}

We consider a 1-to-K communication scenario, where a source transmits private messages to K receivers through a broadcast erasure channel, and the receivers feed back strictly causally and publicly their channel states after each transmission. We explore the achievable rate region when we require that the message to each receiver remains secret - in the information theoretical sense - from all the other receivers. We characterize the capacity of secure communication in all the cases where the capacity of the 1-to-K communication scenario without the requirement of security is known. As a special case, we characterize the secret-message capacity of a single receiver point-to-point erasure channel with public state-feedback in the presence of a passive eavesdropper. We find that in all cases where we have an exact characterization, we can achieve the capacity by using linear complexity two-phase schemes: in the first phase we create appropriate secret keys, and in the second phase we use them to encrypt each message. We find that the amount of key we need is smaller than the size of the message, and equal to the amount of encrypted message the potential eavesdroppers jointly collect. Moreover, we prove that a dishonest receiver that provides deceptive feedback cannot diminish the rate experienced by the honest receivers. We also develop a converse proof which reflects the two-phase structure of our achievability scheme. As a side result, our technique leads to a new outer bound proof for the non-secure communication problem.

@article{DBLP:journals/tit/KarakusWD15,
 abstract = {We investigate how to exploit intermittent feedback for interference management by studying the two-user Gaussian interference channel (IC). We approximately characterize (within a universal constant) the capacity region for the Gaussian IC with intermittent feedback. We exactly characterize the the capacity region of the linear deterministic version of the problem, which gives us insight into the Gaussian problem. We find that the characterization only depends on the forward channel parameters and the marginal probability distribution of each feedback link. The result shows that passive and unreliable feedback can be harnessed to provide multiplicative capacity gain in Gaussian interference channels. We find that when the feedback links are active with sufficiently large probabilities, the perfect feedback sum-capacity is achieved to within a constant gap. In contrast to other schemes developed for interference channel with feedback, our achievable scheme makes use of quantize-map-and-forward to relay the information obtained through feedback, performs forward decoding, and does not use structured codes. We also develop new outer bounds enabling us to obtain the (approximate) characterization of the capacity region.},
 author = {Can Karakus and
I-Hsiang Wang and
Suhas N. Diggavi},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/journals/tit/KarakusWD15},
 doi = {10.1109/TIT.2015.2452915},
 file = {:papers:kwd_it15.pdf},
 journal = {IEEE Transactions on Information Theory},
 number = {9},
 pages = {4663--4699},
 tags = {journal,WiIntMgmt,IT,WiNet,WNIF},
 timestamp = {Thu, 20 Aug 2015 14:54:38 +0200},
 title = {Gaussian Interference Channel With Intermittent Feedback},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1408.4590},
 volume = {61},
 year = {2015}
}

We investigate how to exploit intermittent feedback for interference management by studying the two-user Gaussian interference channel (IC). We approximately characterize (within a universal constant) the capacity region for the Gaussian IC with intermittent feedback. We exactly characterize the the capacity region of the linear deterministic version of the problem, which gives us insight into the Gaussian problem. We find that the characterization only depends on the forward channel parameters and the marginal probability distribution of each feedback link. The result shows that passive and unreliable feedback can be harnessed to provide multiplicative capacity gain in Gaussian interference channels. We find that when the feedback links are active with sufficiently large probabilities, the perfect feedback sum-capacity is achieved to within a constant gap. In contrast to other schemes developed for interference channel with feedback, our achievable scheme makes use of quantize-map-and-forward to relay the information obtained through feedback, performs forward decoding, and does not use structured codes. We also develop new outer bounds enabling us to obtain the (approximate) characterization of the capacity region.

@article{DBLP:journals/tit/KolteOD15,
 abstract = {We study a simple question: when are dynamic relaying strategies essential in optimizing the diversity-multiplexing tradeoff (DMT) in half-duplex wireless relay networks? This is motivated by apparently two contrasting results even for a simple 3 node network, with a single half-duplex relay. When all channels are assumed to be i.i.d. fading, a static schedule where the relay listens half the time and transmits half the time combined with quantize-map-forward (QMF) relaying is known to achieve the full-duplex performance. However, when there is no direct link between source and destination, a dynamic-decode-forward (DDF) strategy is needed to achieve the optimal tradeoff. In this case, a static schedule is strictly suboptimal and the optimal tradeoff is significantly worse than the full-duplex performance. In this paper we study the general case when the direct link is neither as strong as the other links nor fully non-existent, and identify regimes where dynamic schedules are necessary and those where static schedules are enough. We identify 4 qualitatively different regimes for the single relay channel where the tradeoff between diversity and multiplexing is significantly different. We show that in all these regimes one of the above two strategies is sufficient to achieve the optimal tradeoff by developing a new upper bound on the best achievable tradeoff under channel state information available only at the receivers. A natural next question is whether these two strategies are sufficient to achieve the DMT of more general half-duplex wireless networks. We propose a generalization of the two existing schemes through a dynamic QMF (DQMF) strategy, where the relay listens for a fraction of time depending on received CSI but not long enough to be able to decode. We show that such a DQMF strategy is needed to achieve the optimal DMT in a parallel channel with two relays.},
 author = {Ritesh Kolte and
Ayfer {\"{O}}zg{\"{u}}r and
Suhas N. Diggavi},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/journals/tit/KolteOD15},
 doi = {10.1109/TIT.2015.2402163},
 file = {:papers:KolteAD_IT15.pdf},
 journal = {IEEE Transactions on Information Theory},
 number = {4},
 pages = {1720--1738},
 tags = {journal,IT,WiNet},
 timestamp = {Thu, 20 Aug 2015 14:54:38 +0200},
 title = {When Are Dynamic Relaying Strategies Necessary in Half-Duplex Wireless Networks?},
 type = {2},
 url_arxiv = {https://arxiv.org/abs/1405.1782},
 volume = {61},
 year = {2015}
}

We study a simple question: when are dynamic relaying strategies essential in optimizing the diversity-multiplexing tradeoff (DMT) in half-duplex wireless relay networks? This is motivated by apparently two contrasting results even for a simple 3 node network, with a single half-duplex relay. When all channels are assumed to be i.i.d. fading, a static schedule where the relay listens half the time and transmits half the time combined with quantize-map-forward (QMF) relaying is known to achieve the full-duplex performance. However, when there is no direct link between source and destination, a dynamic-decode-forward (DDF) strategy is needed to achieve the optimal tradeoff. In this case, a static schedule is strictly suboptimal and the optimal tradeoff is significantly worse than the full-duplex performance. In this paper we study the general case when the direct link is neither as strong as the other links nor fully non-existent, and identify regimes where dynamic schedules are necessary and those where static schedules are enough. We identify 4 qualitatively different regimes for the single relay channel where the tradeoff between diversity and multiplexing is significantly different. We show that in all these regimes one of the above two strategies is sufficient to achieve the optimal tradeoff by developing a new upper bound on the best achievable tradeoff under channel state information available only at the receivers. A natural next question is whether these two strategies are sufficient to achieve the DMT of more general half-duplex wireless networks. We propose a generalization of the two existing schemes through a dynamic QMF (DQMF) strategy, where the relay listens for a fraction of time depending on received CSI but not long enough to be able to decode. We show that such a DQMF strategy is needed to achieve the optimal DMT in a parallel channel with two relays.

@article{kolte2015dynamic,
 author = {Kolte, Ritesh and {\"O}zg{\"u}r, Ayfer and Diggavi, Suhas},
 journal = {IEEE Transactions on Information Theory},
 number = {4},
 pages = {1720--1738},
 publisher = {IEEE},
 tags = {journal,WiNet,WNIF},
 title = {When are dynamic relaying strategies necessary in half-duplex wireless networks?},
 type = {2},
 volume = {61},
 year = {2015}
}

@inproceedings{7218445,
 abstract = {Emerging heterogeneous wireless architectures consist of a dense deployment of local-coverage wireless access points (APs) with high data rates, along with sparsely-distributed, large-coverage macro-cell base stations (BS). We design a coded caching-and-delivery scheme for such architectures that equips APs with storage, enabling content pre-fetching prior to knowing user demands. Users requesting content are served by connecting to local APs with cached content, as well as by listening to a BS broadcast transmission. For any given content popularity profile, the goal is to design the caching-and-delivery scheme so as to optimally trade off the transmission cost at the BS against the storage cost at the APs and the user cost of connecting to multiple APs. We design a coded caching scheme for non-uniform content popularity that dynamically allocates user access to APs based on requested content. We demonstrate the approximate optimality of our scheme with respect to information-theoretic bounds. We numerically evaluate it on a YouTube dataset and quantify the trade-off between transmission rate, storage, and access cost. Our numerical results also suggest the intriguing possibility that, to gain most of the benefits of coded caching, it suffices to divide the content into a small number of popularity classes.},
 author = {Hachem, J. and Karamchandani, N. and Diggavi, S.},
 booktitle = {Computer Communications (INFOCOM), 2015 IEEE Conference on},
 month = {April},
 pages = {756-764},
 tags = {conf,Caching,IT,WiNet},
 title = {Content caching and delivery over heterogeneous wireless networks},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1404.6560},
 year = {2015}
}

Emerging heterogeneous wireless architectures consist of a dense deployment of local-coverage wireless access points (APs) with high data rates, along with sparsely-distributed, large-coverage macro-cell base stations (BS). We design a coded caching-and-delivery scheme for such architectures that equips APs with storage, enabling content pre-fetching prior to knowing user demands. Users requesting content are served by connecting to local APs with cached content, as well as by listening to a BS broadcast transmission. For any given content popularity profile, the goal is to design the caching-and-delivery scheme so as to optimally trade off the transmission cost at the BS against the storage cost at the APs and the user cost of connecting to multiple APs. We design a coded caching scheme for non-uniform content popularity that dynamically allocates user access to APs based on requested content. We demonstrate the approximate optimality of our scheme with respect to information-theoretic bounds. We numerically evaluate it on a YouTube dataset and quantify the trade-off between transmission rate, storage, and access cost. Our numerical results also suggest the intriguing possibility that, to gain most of the benefits of coded caching, it suffices to divide the content into a small number of popularity classes.

@inproceedings{7282746,
 abstract = {It has been recently established that joint design of content delivery and storage (coded caching) can significantly improve performance over conventional caching. This has also been extended to the case when content has non-uniform popularity through several models. In this paper we focus on a multi-level popularity model, where content is divided into levels based on popularity. We consider two extreme cases of user distribution across caches for the multi-level popularity model: a single user per cache (single-user setup) versus a large number of users per cache (multi-user setup). When the capacity approximation is universal (independent of number of popularity levels as well as number of users, files and caches), we demonstrate a dichotomy in the order-optimal strategies for these two extreme cases. In the multi-user case, sharing memory among the levels is order-optimal, whereas for the single-user case clustering popularity levels and allocating all the memory to them is the order-optimal scheme. In proving these results, we develop new information-theoretic lower bounds for the problem.},
 author = {Hachem, J. and Karamchandani, N. and Diggavi, S.},
 booktitle = {Information Theory (ISIT), 2015 IEEE International Symposium on},
 month = {June},
 pages = {1701-1705},
 tags = {conf,Caching,IT,WiNet},
 title = {Effect of number of users in multi-level coded caching},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1504.05931},
 year = {2015}
}

It has been recently established that joint design of content delivery and storage (coded caching) can significantly improve performance over conventional caching. This has also been extended to the case when content has non-uniform popularity through several models. In this paper we focus on a multi-level popularity model, where content is divided into levels based on popularity. We consider two extreme cases of user distribution across caches for the multi-level popularity model: a single user per cache (single-user setup) versus a large number of users per cache (multi-user setup). When the capacity approximation is universal (independent of number of popularity levels as well as number of users, files and caches), we demonstrate a dichotomy in the order-optimal strategies for these two extreme cases. In the multi-user case, sharing memory among the levels is order-optimal, whereas for the single-user case clustering popularity levels and allocating all the memory to them is the order-optimal scheme. In proving these results, we develop new information-theoretic lower bounds for the problem.

@inproceedings{7282993,
 author = {Mishra, S. and Shoukry, Y. and Karamchandani, N. and Diggavi, S. and Tabuada, P.},
 booktitle = {Information Theory (ISIT), 2015 IEEE International Symposium on},
 file = {:papers:sse_mskdt.pdf},
 month = {June},
 pages = {2929-2933},
 tags = {conf,IT,SecureCPS},
 title = {Secure state estimation: Optimal guarantees against sensor attacks in the presence of noise},
 type = {4},
 year = {2015}
}

@inproceedings{DBLP:conf/isit/BaiDS15,
 author = {Jingwen Bai and
Suhas N. Diggavi and
Ashutosh Sabharwal},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/conf/isit/BaiDS15},
 booktitle = {IEEE International Symposium on Information Theory, ISIT 2015,
Hong Kong, China, June 14-19, 2015},
 crossref = {DBLP:conf/isit/2015},
 doi = {10.1109/ISIT.2015.7282578},
 file = {:papers:bds_fd.pdf},
 pages = {864--868},
 tags = {conf,WiIntMgmt,IT,WiNet},
 timestamp = {Mon, 12 Oct 2015 09:21:34 +0200},
 title = {On degrees-of-freedom of multi-user MIMO full-duplex network},
 type = {4},
 year = {2015}
}

@inproceedings{DBLP:conf/isit/KarakusD15,
 author = {Can Karakus and
Suhas N. Diggavi},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/conf/isit/KarakusD15},
 booktitle = {IEEE International Symposium on Information Theory, ISIT 2015,
Hong Kong, China, June 14-19, 2015},
 crossref = {DBLP:conf/isit/2015},
 doi = {10.1109/ISIT.2015.7282609},
 file = {:papers:isit15_fd_kd.pdf},
 pages = {1019--1023},
 tags = {conf,WiIntMgmt,IT,WiNet},
 timestamp = {Mon, 12 Oct 2015 09:21:34 +0200},
 title = {Opportunistic scheduling for full-duplex uplink-downlink networks},
 type = {4},
 year = {2015}
}

@inproceedings{DBLP:conf/isit/MishraDPD15,
 abstract = {We study oblivious transfer (OT) between Alice and Bob in the presence of an eavesdropper Eve over a degraded wiretapped binary erasure channel from Alice to Bob and Eve. In addition to the privacy goals of oblivious transfer between Alice and Bob, we require privacy of Alice and Bob's private data from Eve. In previous work we derived the OT capacity (in the honest-but-curious model) of the wiretapped binary independent erasure channel where the erasure processes of Bob and Eve are independent. Here we derive a lower bound on the OT capacity in the same secrecy model when the wiretapped binary erasure channel is degraded in favour of Bob.},
 author = {Manoj Mishra and
Bikash Kumar Dey and
Vinod M. Prabhakaran and
Suhas N. Diggavi},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/conf/isit/MishraDPD15},
 booktitle = {IEEE International Symposium on Information Theory, ISIT 2015,
Hong Kong, China, June 14-19, 2015},
 crossref = {DBLP:conf/isit/2015},
 doi = {10.1109/ISIT.2015.7282639},
 file = {:papers:isit15_mdpd.pdf},
 pages = {1169--1173},
 tags = {conf,IT},
 timestamp = {Mon, 12 Oct 2015 09:21:34 +0200},
 title = {On the oblivious transfer capacity of the degraded wiretapped binary erasure channel},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/1504.04499},
 year = {2015}
}

We study oblivious transfer (OT) between Alice and Bob in the presence of an eavesdropper Eve over a degraded wiretapped binary erasure channel from Alice to Bob and Eve. In addition to the privacy goals of oblivious transfer between Alice and Bob, we require privacy of Alice and Bob's private data from Eve. In previous work we derived the OT capacity (in the honest-but-curious model) of the wiretapped binary independent erasure channel where the erasure processes of Bob and Eve are independent. Here we derive a lower bound on the OT capacity in the same secrecy model when the wiretapped binary erasure channel is degraded in favour of Bob.

@inproceedings{DBLP:conf/isit/TianCDS15,
 abstract = {We investigate whether uncoded schemes are optimal for Gaussian sources on multiuser Gaussian channels. Particularly, we consider two problems: the first is to send correlated Gaussian sources on a Gaussian broadcast channel where each receiver is interested in reconstructing only one source component (or one specific linear function of the sources) under the mean squared error distortion measure; the second is to send correlated Gaussian sources on a Gaussian multiple-access channel, where each transmitter observes a noisy combination of the source, and the receiver wishes to reconstruct the individual source components (or individual linear functions) under the mean squared error distortion measure. It is shown that when the channel parameters match certain general conditions, the induced distortion tuples are on the boundary of the achievable distortion region, and thus optimal. Instead of following the conventional approach of attempting to characterize the achievable distortion region, we ask the question whether and how a match can be effectively determined. This decision problem formulation helps to circumvent the difficult optimization problem often embedded in region characterization problems, and it also leads us to focus on the critical conditions in the outer bounds that make the inequalities become equalities, which effectively decouple the overall problem into several simpler sub-problems. Optimality results previously unknown in the literature are obtained using this novel approach. As a byproduct of the investigation, novel outer bounds are derived for these two problems.},
 author = {Chao Tian and
Jun Chen and
Suhas N. Diggavi and
Shlomo Shamai},
 bibsource = {dblp computer science bibliography, http://dblp.org},
 biburl = {http://dblp.uni-trier.de/rec/bib/conf/isit/TianCDS15},
 booktitle = {IEEE International Symposium on Information Theory, ISIT 2015,
Hong Kong, China, June 14-19, 2015},
 crossref = {DBLP:conf/isit/2015},
 doi = {10.1109/ISIT.2015.7282500},
 pages = {476--480},
 tags = {conf, IT},
 timestamp = {Mon, 12 Oct 2015 09:21:34 +0200},
 title = {Matched multiuser Gaussian source-channel communications via uncoded
schemes},
 type = {4},
 url_arxiv = {https://arxiv.org/abs/1502.05428},
 year = {2015}
}

We investigate whether uncoded schemes are optimal for Gaussian sources on multiuser Gaussian channels. Particularly, we consider two problems: the first is to send correlated Gaussian sources on a Gaussian broadcast channel where each receiver is interested in reconstructing only one source component (or one specific linear function of the sources) under the mean squared error distortion measure; the second is to send correlated Gaussian sources on a Gaussian multiple-access channel, where each transmitter observes a noisy combination of the source, and the receiver wishes to reconstruct the individual source components (or individual linear functions) under the mean squared error distortion measure. It is shown that when the channel parameters match certain general conditions, the induced distortion tuples are on the boundary of the achievable distortion region, and thus optimal. Instead of following the conventional approach of attempting to characterize the achievable distortion region, we ask the question whether and how a match can be effectively determined. This decision problem formulation helps to circumvent the difficult optimization problem often embedded in region characterization problems, and it also leads us to focus on the critical conditions in the outer bounds that make the inequalities become equalities, which effectively decouple the overall problem into several simpler sub-problems. Optimality results previously unknown in the literature are obtained using this novel approach. As a byproduct of the investigation, novel outer bounds are derived for these two problems.

@inproceedings{mishra2015oblivious,
 author = {Mishra, Manoj and Dey, Bikash Kumar and Prabhakaran, Vinod M and Diggavi, Suhas},
 booktitle = {2015 IEEE International Symposium on Information Theory (ISIT)},
 organization = {IEEE},
 pages = {1169--1173},
 tags = {conf,IT,WNS},
 title = {On the oblivious transfer capacity of the degraded wiretapped binary erasure channel},
 type = {4},
 year = {2015}
}

@inproceedings{PyCRA,
 author = {Shoukry, Y. and Martin, P. and Yona, Y. and Diggavi, S. and Srivastava, M.},
 booktitle = {The 22nd ACM Conference on Computer and Communications Security (CCS 2015)},
 file = {:papers:pycra.pdf},
 month = {October},
 pages = {1004-1015},
 tags = {conf,IT,SecureCPS,Hardware},
 title = {PyCRA: Physical Challenge-Response Authentication For Active Sensors Under Spoofing Attacks},
 type = {4},
 year = {2015}
}

@inproceedings{sebastianKDW15,
 author = {Sebastian, J. and Karakus, C. and Diggavi, S. and Wang, I.-H.},
 booktitle = {Allerton Conference on Communication, Control and Computing, 2015},
 file = {:papers:allerton15_fading.pdf},
 month = {October},
 tags = {conf,WiIntMgmt,IT,WiNet},
 title = {Rate splitting is approximately optimal for fading gaussian interference channels},
 type = {4},
 year = {2015}
}

@article{avestimehr2015approximation,
 author = {Avestimehr, A Salman and Diggavi, Suhas N and Tian, Chao and Tse, David NC},
 journal = {Foundations and Trends in Communications and Information Theory},
 number = {1-2},
 pages = {1--183},
 publisher = {Now Publishers Inc. Hanover, MA, USA},
 tags = {monograph,ANIT},
 title = {An approximation approach to network information theory},
 type = {5},
 volume = {12},
 year = {2015}
}

@article{6671968,
 abstract = {We consider the source-channel separation architecture for lossy source coding in communication networks. It is shown that the separation approach is optimal in two general scenarios and is approximately optimal in a third scenario. The two scenarios for which separation is optimal complement each other: the first is when the memoryless sources at source nodes are arbitrarily correlated, each of which is to be reconstructed at possibly multiple destinations within certain distortions, but the channels in this network are synchronized, orthogonal, and memoryless point-to-point channels; the second is when the memoryless sources are mutually independent, each of which is to be reconstructed only at one destination within a certain distortion, but the channels are general, including multi-user channels, such as multiple access, broadcast, interference, and relay channels, possibly with feedback. The third scenario, for which we demonstrate approximate optimality of source-channel separation, generalizes the second scenario by allowing each source to be reconstructed at multiple destinations with different distortions. For this case, the loss from optimality using the separation approach can be upper-bounded when a difference distortion measure is taken, and in the special case of quadratic distortion measure, this leads to universal constant bounds.},
 author = {Chao Tian and Jun Chen and Diggavi, S.N. and Shamai, S.},
 doi = {10.1109/TIT.2013.2291787},
 issn = {0018-9448},
 journal = {Information Theory, IEEE Transactions on},
 month = {Feb},
 number = {2},
 pages = {904-918},
 tags = {journal,SrcChan,IT},
 title = {Optimality and Approximate Optimality of Source-Channel Separation in Networks},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1004.2648},
 volume = {60},
 year = {2014}
}

We consider the source-channel separation architecture for lossy source coding in communication networks. It is shown that the separation approach is optimal in two general scenarios and is approximately optimal in a third scenario. The two scenarios for which separation is optimal complement each other: the first is when the memoryless sources at source nodes are arbitrarily correlated, each of which is to be reconstructed at possibly multiple destinations within certain distortions, but the channels in this network are synchronized, orthogonal, and memoryless point-to-point channels; the second is when the memoryless sources are mutually independent, each of which is to be reconstructed only at one destination within a certain distortion, but the channels are general, including multi-user channels, such as multiple access, broadcast, interference, and relay channels, possibly with feedback. The third scenario, for which we demonstrate approximate optimality of source-channel separation, generalizes the second scenario by allowing each source to be reconstructed at multiple destinations with different distortions. For this case, the loss from optimality using the separation approach can be upper-bounded when a difference distortion measure is taken, and in the special case of quadratic distortion measure, this leads to universal constant bounds.

@article{6727407,
 abstract = {The vast majority of today's critical infrastructure is supported by numerous feedback control loops and an attack on these control loops can have disastrous consequences. This is a major concern since modern control systems are becoming large and decentralized and thus more vulnerable to attacks. This paper is concerned with the estimation and control of linear systems when some of the sensors or actuators are corrupted by an attacker. We give a new simple characterization of the maximum number of attacks that can be detected and corrected as a function of the pair (A,C) of the system and we show in particular that it is impossible to accurately reconstruct the state of a system if more than half the sensors are attacked. In addition, we show how the design of a secure local control loop can improve the resilience of the system. When the number of attacks is smaller than a threshold, we propose an efficient algorithm inspired from techniques in compressed sensing to estimate the state of the plant despite attacks. We give a theoretical characterization of the performance of this algorithm and we show on numerical simulations that the method is promising and allows to reconstruct the state accurately despite attacks. Finally, we consider the problem of designing output-feedback controllers that stabilize the system despite sensor attacks. We show that a principle of separation between estimation and control holds and that the design of resilient output feedback controllers can be reduced to the design of resilient state estimators.},
 author = {Fawzi, H. and Tabuada, P. and Diggavi, S.},
 doi = {10.1109/TAC.2014.2303233},
 issn = {0018-9286},
 journal = {Automatic Control, IEEE Transactions on},
 month = {June},
 number = {6},
 pages = {1454-1467},
 tags = {journal,SecureCPS},
 title = {Secure Estimation and Control for Cyber-Physical Systems Under Adversarial Attacks},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1205.5073},
 volume = {59},
 year = {2014}
}

The vast majority of today's critical infrastructure is supported by numerous feedback control loops and an attack on these control loops can have disastrous consequences. This is a major concern since modern control systems are becoming large and decentralized and thus more vulnerable to attacks. This paper is concerned with the estimation and control of linear systems when some of the sensors or actuators are corrupted by an attacker. We give a new simple characterization of the maximum number of attacks that can be detected and corrected as a function of the pair (A,C) of the system and we show in particular that it is impossible to accurately reconstruct the state of a system if more than half the sensors are attacked. In addition, we show how the design of a secure local control loop can improve the resilience of the system. When the number of attacks is smaller than a threshold, we propose an efficient algorithm inspired from techniques in compressed sensing to estimate the state of the plant despite attacks. We give a theoretical characterization of the performance of this algorithm and we show on numerical simulations that the method is promising and allows to reconstruct the state accurately despite attacks. Finally, we consider the problem of designing output-feedback controllers that stabilize the system despite sensor attacks. We show that a principle of separation between estimation and control holds and that the design of resilient output feedback controllers can be reduced to the design of resilient state estimators.

@article{TDGj09,
 abstract = {Wireless network topologies change over time and maintaining routes requires frequent updates. Updates are
costly in terms of consuming throughput available for data transmission, which is precious in wireless networks.
In this paper, we ask whether there exist low-overhead schemes that produce near-optimal routes. We show that random geometric graphs and variants have a nice property that they
are doubling metric spaces, with high probability. This enables us to design hierarchical routing schemes with provable performance that account for the user mobility and scale well with network size.},
 author = {D. Tschopp and S N. Diggavi and M. Grossglauser},
 file = {:papers:tdgrouting09.pdf},
 journal = {Random Structures and Algorithms},
 label = {tdg_j09},
 month = {July},
 note = {},
 pages = {},
 tags = {journal,WiNet,CSalgo,DynamWiNet,MetricEmb,WiNetRouting},
 title = {Hierarchical Routing over Dynamic Wireless Networks},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0909.2504},
 volume = {},
 year = {2014}
}

Wireless network topologies change over time and maintaining routes requires frequent updates. Updates are costly in terms of consuming throughput available for data transmission, which is precious in wireless networks. In this paper, we ask whether there exist low-overhead schemes that produce near-optimal routes. We show that random geometric graphs and variants have a nice property that they are doubling metric spaces, with high probability. This enables us to design hierarchical routing schemes with provable performance that account for the user mobility and scale well with network size.

@inproceedings{6848197,
 abstract = {Physical layer cooperation of a source with a relay can significantly boost the performance of a wireless connection. However, the best practical relaying scheme can vary depending on the relative strengths of the channels that connect the source, relay and destination. This paper proposes and evaluates QUILT, a system for physical-layer relaying that seamlessly adapts to the underlying network configuration to achieve competitive or better performance as compared to the best current approaches. QUILT combines on-demand, opportunistic use of Decode-Forward (DF) or Quantize-Map-Forward (QMF) followed by interleaving at the relay, with hybrid decoding at the destination that extracts information from received frames even if these are not decodable. We theoretically quantify how our design choices for QUILT affect the system performance. We also deploy QUILT on the WarpLab software radio platform, and show through over-the-air experiments up to 5 times FER improvement over the next best cooperative protocol.},
 author = {Brahma, S. and Duarte, M. and Sengupta, A and I-Hsiang Wang and Fragouli, C. and Diggavi, S.},
 booktitle = {INFOCOM, 2014 Proceedings IEEE},
 doi = {10.1109/INFOCOM.2014.6848197},
 file = {:papers:quilt.pdf},
 month = {April},
 pages = {2508-2516},
 tags = {conf,WiNet,WiNetInfFlow},
 title = {QUILT: A Decode/Quantize-Interleave-Transmit approach to cooperative relaying},
 type = {4},
 year = {2014}
}

Physical layer cooperation of a source with a relay can significantly boost the performance of a wireless connection. However, the best practical relaying scheme can vary depending on the relative strengths of the channels that connect the source, relay and destination. This paper proposes and evaluates QUILT, a system for physical-layer relaying that seamlessly adapts to the underlying network configuration to achieve competitive or better performance as compared to the best current approaches. QUILT combines on-demand, opportunistic use of Decode-Forward (DF) or Quantize-Map-Forward (QMF) followed by interleaving at the relay, with hybrid decoding at the destination that extracts information from received frames even if these are not decodable. We theoretically quantify how our design choices for QUILT affect the system performance. We also deploy QUILT on the WarpLab software radio platform, and show through over-the-air experiments up to 5 times FER improvement over the next best cooperative protocol.

@inproceedings{6874794,
 abstract = {Recent work has demonstrated that, for content caching, joint design of storage and delivery can yield significant benefits over conventional caching approaches. This is based on storing content in the caches in a way that creates coded-multicast opportunities even among users with different demands. Such a coded-caching scheme has been shown to be order-optimal for a caching system with single-level content, i.e., one where all content is uniformly popular. In this work, we consider a system with content divided into multiple levels, based on varying degrees of popularity. The main contribution of this work is the derivation of an information-theoretic outer bound for the multi-level setup, and the demonstration that, under some natural regularity conditions, a memory-sharing scheme, which operates each level in isolation according to a single-level coded caching scheme, is in fact order-optimal with respect to this outer bound.},
 author = {Hachem, J. and Karamchandani, N. and Diggavi, S.},
 booktitle = {Information Theory (ISIT), 2014 IEEE International Symposium on},
 doi = {10.1109/ISIT.2014.6874794},
 month = {June},
 pages = {56-60},
 tags = {conf,Caching,IT,WiNet},
 title = {Multi-level coded caching},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1404.6563},
 year = {2014}
}

Recent work has demonstrated that, for content caching, joint design of storage and delivery can yield significant benefits over conventional caching approaches. This is based on storing content in the caches in a way that creates coded-multicast opportunities even among users with different demands. Such a coded-caching scheme has been shown to be order-optimal for a caching system with single-level content, i.e., one where all content is uniformly popular. In this work, we consider a system with content divided into multiple levels, based on varying degrees of popularity. The main contribution of this work is the derivation of an information-theoretic outer bound for the multi-level setup, and the demonstration that, under some natural regularity conditions, a memory-sharing scheme, which operates each level in isolation according to a single-level coded caching scheme, is in fact order-optimal with respect to this outer bound.

@inproceedings{6874917,
 abstract = {We study parallel symmetric 2-user interference channels when the interference is bursty and feedback is available from the respective receivers. Presence of interference in each subcarrier is modeled as a memoryless Bernoulli random state. The states across subcarriers are drawn from an arbitrary joint distribution with the same marginal probability for each subcarrier and instantiated i.i.d. over time. For the linear deterministic setup, we give a complete characterization of the capacity region. For the setup with Gaussian noise, we give outer bounds and a tight generalized degrees of freedom characterization. We propose a novel helping mechanism which enables subcarriers in very strong interference regime to help in recovering interfered signals for subcarriers in strong and weak interference regimes. Depending on the interference and burstiness regime, the inner bounds either employ the proposed helping mechanism to code across subcarriers or treat the subcarriers separately. The outer bounds demonstrate a connection to a subset entropy inequality by Madiman and Tetali [4].},
 author = {Mishra, S. and I-Hsiang Wang and Diggavi, S.},
 booktitle = {Information Theory (ISIT), 2014 IEEE International Symposium on},
 doi = {10.1109/ISIT.2014.6874917},
 month = {June},
 pages = {671-675},
 tags = {conf,WiIntMgmt,IT,WNIF},
 title = {Harnessing bursty interference in multicarrier systems with feedback},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1404.6044},
 year = {2014}
}

We study parallel symmetric 2-user interference channels when the interference is bursty and feedback is available from the respective receivers. Presence of interference in each subcarrier is modeled as a memoryless Bernoulli random state. The states across subcarriers are drawn from an arbitrary joint distribution with the same marginal probability for each subcarrier and instantiated i.i.d. over time. For the linear deterministic setup, we give a complete characterization of the capacity region. For the setup with Gaussian noise, we give outer bounds and a tight generalized degrees of freedom characterization. We propose a novel helping mechanism which enables subcarriers in very strong interference regime to help in recovering interfered signals for subcarriers in strong and weak interference regimes. Depending on the interference and burstiness regime, the inner bounds either employ the proposed helping mechanism to code across subcarriers or treat the subcarriers separately. The outer bounds demonstrate a connection to a subset entropy inequality by Madiman and Tetali [4].

@inproceedings{6874939,
 abstract = {We characterize the secret message capacity of the triangle network, that consists of a source, a relay and a destination connected through orthogonal erasure channels. A passive eavesdropper, Eve, wiretaps any one of the three channels. The source and the relay can each generate unlimited private randomness; the relay and the destination can publicly provide strictly causal channel state information. Our achievable scheme is expressed through a linear program (LP) with 11 inequalities that captures a minimal set of secret key generation methods and the use of them for message encryption. Our outer bound is expressed also through a linear program, in this case with 41 constraints, constructed from general information inequalities. We prove that the optimal value of the outer bound LP is no larger than that of the scheme LP, which implies that the solution of the achievable scheme LP is the capacity. We find that equipping the relay with private randomness increases the secrecy rate by more than 40\% in some cases and that cut-set bounds, directly applied in the network, are not always tight. Because the derivation of the inner and outer bound are both lengthy, we describe in this paper the achievability scheme, outline the outer bound, and provide the full derivations online [1]. We also make available Matlab functions that take as input the erasure probabilities and evaluate the inner and outer bounds.},
 author = {Czap, L. and Prabhakaran, V.M. and Diggavi, S. and Fragouli, C.},
 booktitle = {Information Theory (ISIT), 2014 IEEE International Symposium on},
 doi = {10.1109/ISIT.2014.6874939},
 file = {:papers:tri_net_sec.pdf},
 month = {June},
 pages = {781-785},
 tags = {conf,WiNetSec,IT},
 title = {Triangle network secrecy},
 type = {4},
 year = {2014}
}

We characterize the secret message capacity of the triangle network, that consists of a source, a relay and a destination connected through orthogonal erasure channels. A passive eavesdropper, Eve, wiretaps any one of the three channels. The source and the relay can each generate unlimited private randomness; the relay and the destination can publicly provide strictly causal channel state information. Our achievable scheme is expressed through a linear program (LP) with 11 inequalities that captures a minimal set of secret key generation methods and the use of them for message encryption. Our outer bound is expressed also through a linear program, in this case with 41 constraints, constructed from general information inequalities. We prove that the optimal value of the outer bound LP is no larger than that of the scheme LP, which implies that the solution of the achievable scheme LP is the capacity. We find that equipping the relay with private randomness increases the secrecy rate by more than 40% in some cases and that cut-set bounds, directly applied in the network, are not always tight. Because the derivation of the inner and outer bound are both lengthy, we describe in this paper the achievability scheme, outline the outer bound, and provide the full derivations online [1]. We also make available Matlab functions that take as input the erasure probabilities and evaluate the inner and outer bounds.

@inproceedings{6875091,
 abstract = {We consider oblivious transfer between Alice and Bob in the presence of an eavesdropper Eve when there is a broadcast channel from Alice to Bob and Eve. In addition to the secrecy constraints of Alice and Bob, Eve should not learn the private data of Alice and Bob. When the broadcast channel consists of two independent binary erasure channels, we derive the oblivious transfer capacity for both 2-privacy (where the eavesdropper may collude with either party) and 1-privacy (where there are no collusions).},
 author = {Mishra, M. and Dey, B.K. and Prabhakaran, V.M. and Diggavi, S.},
 booktitle = {Information Theory (ISIT), 2014 IEEE International Symposium on},
 doi = {10.1109/ISIT.2014.6875091},
 month = {June},
 pages = {1539-1543},
 tags = {conf,IT,WiNetSec},
 title = {The oblivious transfer capacity of the wiretapped binary erasure channel},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1404.6614},
 year = {2014}
}

We consider oblivious transfer between Alice and Bob in the presence of an eavesdropper Eve when there is a broadcast channel from Alice to Bob and Eve. In addition to the secrecy constraints of Alice and Bob, Eve should not learn the private data of Alice and Bob. When the broadcast channel consists of two independent binary erasure channels, we derive the oblivious transfer capacity for both 2-privacy (where the eavesdropper may collude with either party) and 1-privacy (where there are no collusions).

@inproceedings{6875212,
 abstract = {It has recently been demonstrated that for single-layer cache networks, jointly designing caching and delivery can enable significant benefits over conventional caching. This was based on strategically designing the cached content to induce coded multicasting opportunities even among users with different demands and without foreknowledge of the user demands. In this work, we extend this coded caching approach to a multi-hop hierarchical content delivery network with two layers of caches. We propose a new caching scheme that combines two basic approaches. The first approach provides coded multicasting opportunities within each layer (through decoding and forwarding); the second approach provides coded multicasting opportunities across multiple layers (through strategic forwarding without decoding). By striking the right balance between these two approaches, we show that the proposed scheme achieves the optimal communication rates to within a constant multiplicative and additive gap. We further show that there is no tension between the rates in each of the two layers up to the aforementioned gap. Thus, both layers can simultaneously operate at approximately the minimum rate.},
 author = {Karamchandani, N. and Niesen, U. and Maddah-Ali, M.A and Diggavi, S.},
 booktitle = {Information Theory (ISIT), 2014 IEEE International Symposium on},
 doi = {10.1109/ISIT.2014.6875212},
 file = {:papers:hier_caching.pdf},
 month = {June},
 pages = {2142-2146},
 tags = {conf,Caching,IT,WiNet},
 title = {Hierarchical coded caching},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1403.7007},
 year = {2014}
}

It has recently been demonstrated that for single-layer cache networks, jointly designing caching and delivery can enable significant benefits over conventional caching. This was based on strategically designing the cached content to induce coded multicasting opportunities even among users with different demands and without foreknowledge of the user demands. In this work, we extend this coded caching approach to a multi-hop hierarchical content delivery network with two layers of caches. We propose a new caching scheme that combines two basic approaches. The first approach provides coded multicasting opportunities within each layer (through decoding and forwarding); the second approach provides coded multicasting opportunities across multiple layers (through strategic forwarding without decoding). By striking the right balance between these two approaches, we show that the proposed scheme achieves the optimal communication rates to within a constant multiplicative and additive gap. We further show that there is no tension between the rates in each of the two layers up to the aforementioned gap. Thus, both layers can simultaneously operate at approximately the minimum rate.

@inproceedings{chenozgur14,
 abstract = {In this paper we examine the value of feedback that comes from overhearing, without dedicated feedback resources. We focus on a simple model for this purpose: a deterministic two-hop interference channel, where feedback comes from overhearing the forward-links. A new aspect brought by this setup is the dual-role of the relay signal. While the relay signal needs to convey the source message to its corresponding destination, it can also provide a feedback signal which can potentially increase the capacity of the first hop. We derive inner and outer bounds on the sum capacity which match for a large range of the parameter values. Our results identify the parameter ranges where overhearing can provide non-negative capacity gain and can even achieve the performance with dedicated-feedback resources. The results also provide insights into which transmissions are most useful to overhear.},
 author = {Chen, J. and Ozgur, A. and Diggavi, S.N.},
 booktitle = {Annual Allerton Conference on Communication, Control, and Computing (Allerton)},
 file = {:papers:fto_cod.pdf},
 month = {Sep},
 tags = {conf,IT,WiNet,WiNetInfFlow},
 title = {Feedback through overhearing},
 type = {4},
 year = {2014}
}

In this paper we examine the value of feedback that comes from overhearing, without dedicated feedback resources. We focus on a simple model for this purpose: a deterministic two-hop interference channel, where feedback comes from overhearing the forward-links. A new aspect brought by this setup is the dual-role of the relay signal. While the relay signal needs to convey the source message to its corresponding destination, it can also provide a feedback signal which can potentially increase the capacity of the first hop. We derive inner and outer bounds on the sum capacity which match for a large range of the parameter values. Our results identify the parameter ranges where overhearing can provide non-negative capacity gain and can even achieve the performance with dedicated-feedback resources. The results also provide insights into which transmissions are most useful to overhear.

@inproceedings{mishradey14,
 author = {Mishra, M. and Dey, B. and Prabhakaran, V. and Diggavi, S.N.},
 booktitle = {IEEE Information Theory Workshop (ITW) 2014},
 file = {:papers:isit14_mdpd.pdf},
 month = {Nov},
 tags = {conf,IT,WiNetSec},
 title = {On the oblivious transfer capacity region of the binary erasure broadcast channel},
 type = {4},
 year = {2014}
}

@article{6304926,
 abstract = {We consider the Gaussian “diamond” or parallel relay network, in which a source node transmits a message to a destination node with the help of N relays. Even for the symmetric setting, in which the channel gains to the relays are identical and the channel gains from the relays are identical, the capacity of this channel is unknown in general. The best known capacity approximation is up to an additive gap of order N bits and up to a multiplicative gap of order N2, with both gaps independent of the channel gains. In this paper, we approximate the capacity of the symmetric Gaussian N-relay diamond network up to an additive gap of 1.8 bits and up to a multiplicative gap of a factor 14. Both gaps are independent of the channel gains and, unlike the best previously known result, are also independent of the number of relays N in the network. Achievability is based on bursty amplify-and-forward, showing that this simple scheme is uniformly approximately optimal, both in the low-rate as well as in the high-rate regimes. The upper bound on capacity is based on a careful evaluation of the cut-set bound. We also present approximation results for the asymmetric Gaussian N-relay diamond network. In particular, we show that bursty amplify-and-forward combined with optimal relay selection achieves a rate within a factor O(log4(N)) of capacity with preconstant in the order notation independent of the channel gains.},
 author = {Niesen, U. and Diggavi, S.N.},
 doi = {10.1109/TIT.2012.2219154},
 issn = {0018-9448},
 journal = {Information Theory, IEEE Transactions on},
 month = {Feb},
 number = {2},
 pages = {845-859},
 tags = {journal,WiNet,IT,approxIT},
 title = {The Approximate Capacity of the Gaussian n-Relay Diamond Network},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1008.3813},
 volume = {59},
 year = {2013}
}

We consider the Gaussian “diamond” or parallel relay network, in which a source node transmits a message to a destination node with the help of N relays. Even for the symmetric setting, in which the channel gains to the relays are identical and the channel gains from the relays are identical, the capacity of this channel is unknown in general. The best known capacity approximation is up to an additive gap of order N bits and up to a multiplicative gap of order N2, with both gaps independent of the channel gains. In this paper, we approximate the capacity of the symmetric Gaussian N-relay diamond network up to an additive gap of 1.8 bits and up to a multiplicative gap of a factor 14. Both gaps are independent of the channel gains and, unlike the best previously known result, are also independent of the number of relays N in the network. Achievability is based on bursty amplify-and-forward, showing that this simple scheme is uniformly approximately optimal, both in the low-rate as well as in the high-rate regimes. The upper bound on capacity is based on a careful evaluation of the cut-set bound. We also present approximation results for the asymmetric Gaussian N-relay diamond network. In particular, we show that bursty amplify-and-forward combined with optimal relay selection achieves a rate within a factor O(log4(N)) of capacity with preconstant in the order notation independent of the channel gains.

@article{6481621,
 abstract = {We consider the problem of distributed computation of a target function over a two-user deterministic multiple-access channel. If the target and channel functions are matched (i.e., compute the same function), significant performance gains can be obtained by jointly designing the communication and computation tasks. However, in most situations there is mismatch between these two functions. In this work, we analyze the impact of this mismatch on the performance gains achievable with joint communication and computation designs over separation-based designs. We show that for most pairs of target and channel functions there is no such gain, and separation of communication and computation is optimal.},
 author = {Karamchandani, N. and Niesen, U. and Diggavi, S.},
 doi = {10.1109/JSAC.2013.130405},
 issn = {0733-8716},
 journal = {Selected Areas in Communications, IEEE Journal on},
 month = {April},
 number = {4},
 pages = {666-677},
 tags = {journal,IT},
 title = {Computation over Mismatched Channels},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1204.5059},
 volume = {31},
 year = {2013}
}

We consider the problem of distributed computation of a target function over a two-user deterministic multiple-access channel. If the target and channel functions are matched (i.e., compute the same function), significant performance gains can be obtained by jointly designing the communication and computation tasks. However, in most situations there is mismatch between these two functions. In this work, we analyze the impact of this mismatch on the performance gains achievable with joint communication and computation designs over separation-based designs. We show that for most pairs of target and channel functions there is no such gain, and separation of communication and computation is optimal.

@article{6587830,
 abstract = {Recently, a new relaying strategy, quantize-map-and-forward (QMF) scheme, has been demonstrated to approximately achieve (within an additive constant number of bits) the Gaussian relay network capacity, universally, i.e., for arbitrary topologies, channel gains, and SNRs. This was established using Gaussian codebooks for transmission and random mappings at the relays. In this paper, we develop structured lattice codes that implement the QMF strategy. The main result of this paper is that such structured lattice codes can approximately achieve the Gaussian relay network capacity universally, again within an additive constant. In addition, we establish a similar result for half-duplex networks, where we demonstrate that one can approximately achieve the capacity using fixed transmit-receive (TX-RX) schedules for the relays with no transmit power optimization across the different TX-RX states of the network.},
 author = {Ozgur, A and Diggavi, S.N.},
 doi = {10.1109/TIT.2013.2280167},
 file = {:papers:lattice_qmf.pdf},
 issn = {0018-9448},
 journal = {Information Theory, IEEE Transactions on},
 month = {Dec},
 number = {12},
 pages = {8275-8294},
 tags = {journal,approxIT,WiNet,IT,WiNetInfFlow},
 title = {Approximately Achieving Gaussian Relay Network Capacity With Lattice-Based QMF Codes},
 type = {2},
 volume = {59},
 year = {2013}
}

Recently, a new relaying strategy, quantize-map-and-forward (QMF) scheme, has been demonstrated to approximately achieve (within an additive constant number of bits) the Gaussian relay network capacity, universally, i.e., for arbitrary topologies, channel gains, and SNRs. This was established using Gaussian codebooks for transmission and random mappings at the relays. In this paper, we develop structured lattice codes that implement the QMF strategy. The main result of this paper is that such structured lattice codes can approximately achieve the Gaussian relay network capacity universally, again within an additive constant. In addition, we establish a similar result for half-duplex networks, where we demonstrate that one can approximately achieve the capacity using fixed transmit-receive (TX-RX) schedules for the relays with no transmit power optimization across the different TX-RX states of the network.

@inproceedings{6567030,
 abstract = {We consider the problem where a group of wireless nodes, connected to the same broadcast domain, want to create pairwise secrets, in the presence of an adversary Eve, who tries to listen in and steal these secrets. Existing solutions assume that Eve cannot perform certain computations (e.g., large-integer factorization) in useful time. We ask the question: can we solve this problem without assuming anything about Eve's computational capabilities? We propose a simple secret-agreement protocol, where the wireless nodes keep exchanging bits until they have agreed on pairwise secrets that Eve cannot reconstruct with very high probability. Our protocol relies on Eve's limited network presence (the fact that she cannot be located at an arbitrary number of points in the network at the same time), but assumes nothing about her computational capabilities. We formally show that, under standard theoretical assumptions, our protocol is information-theoretically secure (it leaks zero information to Eve about the secrets). Using a small wireless testbed of smart-phones, we provide experimental evidence that it is feasible for 5 nodes to create thousands of secret bits per second, with their secrecy being independent from the adversary's capabilities.},
 author = {Safaka, I and Fragouli, C. and Argyraki, K. and Diggavi, S.},
 booktitle = {INFOCOM, 2013 Proceedings IEEE},
 doi = {10.1109/INFCOM.2013.6567030},
 file = {:papers:exchanging_secrets.pdf},
 issn = {0743-166X},
 month = {April},
 pages = {2265-2273},
 tags = {conf,WiNetSec,IT},
 title = {Exchanging pairwise secrets efficiently},
 type = {4},
 year = {2013}
}

We consider the problem where a group of wireless nodes, connected to the same broadcast domain, want to create pairwise secrets, in the presence of an adversary Eve, who tries to listen in and steal these secrets. Existing solutions assume that Eve cannot perform certain computations (e.g., large-integer factorization) in useful time. We ask the question: can we solve this problem without assuming anything about Eve's computational capabilities? We propose a simple secret-agreement protocol, where the wireless nodes keep exchanging bits until they have agreed on pairwise secrets that Eve cannot reconstruct with very high probability. Our protocol relies on Eve's limited network presence (the fact that she cannot be located at an arbitrary number of points in the network at the same time), but assumes nothing about her computational capabilities. We formally show that, under standard theoretical assumptions, our protocol is information-theoretically secure (it leaks zero information to Eve about the secrets). Using a small wireless testbed of smart-phones, we provide experimental evidence that it is feasible for 5 nodes to create thousands of secret bits per second, with their secrecy being independent from the adversary's capabilities.

@inproceedings{6570819,
 abstract = {Consider a sender, Alice, who wants to transmit private messages to two receivers, Bob and Calvin, using unreliable wireless broadcast transmissions and short public feedback from Bob and Calvin. In [1], we assumed that Bob and Calvin provide honest feedback, and characterized the secure capacity region of the private messages under the requirement that Bob and Calvin do not learn each other's message. In this paper, we assume that Bob (or Calvin) may provide dishonest feedback; or even control the input message distributions, as is commonly assumed in cryptography literature. We characterize the capacity region in the case of dishonest adversaries, as well as an achievable region for the case when the adversary has complete control on the distribution of the messages. We also design polynomial time protocols for both cases, that rely on the use of coding techniques to mix and secure the private messages. As a side result, we define an extended notion of semantic security for this problem and using a similar approach to [2], we show the equivalence of different security notions.},
 author = {Czap, L. and Prabhakaran, V.M. and Diggavi, S. and Fragouli, C.},
 booktitle = {Network Coding (NetCod), 2013 International Symposium on},
 doi = {10.1109/NetCod.2013.6570819},
 file = {:papers:securing_bc.pdf},
 month = {June},
 pages = {1-6},
 tags = {conf,WiNetSec,IT},
 title = {Securing broadcast against dishonest receivers},
 type = {4},
 year = {2013}
}

Consider a sender, Alice, who wants to transmit private messages to two receivers, Bob and Calvin, using unreliable wireless broadcast transmissions and short public feedback from Bob and Calvin. In [1], we assumed that Bob and Calvin provide honest feedback, and characterized the secure capacity region of the private messages under the requirement that Bob and Calvin do not learn each other's message. In this paper, we assume that Bob (or Calvin) may provide dishonest feedback; or even control the input message distributions, as is commonly assumed in cryptography literature. We characterize the capacity region in the case of dishonest adversaries, as well as an achievable region for the case when the adversary has complete control on the distribution of the messages. We also design polynomial time protocols for both cases, that rely on the use of coding techniques to mix and secure the private messages. As a side result, we define an extended notion of semantic security for this problem and using a similar approach to [2], we show the equivalence of different security notions.

@inproceedings{6612041,
 abstract = {In this paper we study interference management in wireless networks with bursty user traffic. In each time slot, whether a user is on or off for transmission is governed by its own Bernoulli random state. At each transmitter, the states of activities of other users are only available via feedback. We investigate a canonical two-user bursty Gaussian interference channel (IC) with three different feedback models: (1) no feedback, (2) delayed state feedback, and (3) channel output feedback. In all three cases, we characterize the capacity region of the bursty Gaussian IC to within a bounded gap. It turns out that the near-optimal transmit strategies in the non-bursty IC suffice to establish the approximate characterization of capacity in all three cases. In other words, traffic burstiness does not change the high-SNR optimality of the schemes as long as receivers keep track of user activities. Moreover, the capacity region with delayed state feedback is within a bounded gap to that without feedback, and therefore delayed state feedback does not provide significant improvement at high SNR.},
 author = {I-Hsiang Wang and Diggavi, S.},
 booktitle = {Signal Processing Advances in Wireless Communications (SPAWC), 2013 IEEE 14th Workshop on},
 doi = {10.1109/SPAWC.2013.6612041},
 file = {:papers:bursty_ic_spawc.pdf},
 issn = {1948-3244},
 month = {June},
 pages = {205-209},
 tags = {conf,WiIntMgmt,IT,WiNet},
 title = {Interference channels with bursty traffic and delayed feedback},
 type = {4},
 year = {2013}
}

In this paper we study interference management in wireless networks with bursty user traffic. In each time slot, whether a user is on or off for transmission is governed by its own Bernoulli random state. At each transmitter, the states of activities of other users are only available via feedback. We investigate a canonical two-user bursty Gaussian interference channel (IC) with three different feedback models: (1) no feedback, (2) delayed state feedback, and (3) channel output feedback. In all three cases, we characterize the capacity region of the bursty Gaussian IC to within a bounded gap. It turns out that the near-optimal transmit strategies in the non-bursty IC suffice to establish the approximate characterization of capacity in all three cases. In other words, traffic burstiness does not change the high-SNR optimality of the schemes as long as receivers keep track of user activities. Moreover, the capacity region with delayed state feedback is within a bounded gap to that without feedback, and therefore delayed state feedback does not provide significant improvement at high SNR.

@inproceedings{6620180,
 abstract = {We explore the benefit of feedback for physical layer interference management in wireless networks without centralized upper layer control mechanisms. Lack of coordination in the upper layer could make the interference experienced in the physical layer bursty. To understand how to harness such burstiness with feedback, we investigate a two-user bursty interference channel (IC), where the presence of interference is governed by a Bernoulli random state. We completely characterize the capacity region of the symmetric two-user linear deterministic bursty IC with feedback. The proposed two-phase scheme exploits feedback either for refining the previous interfered reception or for relaying additional information to the legitimate receiver of the other user. Matching outer bounds are derived by novel techniques that take the effect of delayed state information into account. We also use insights from the deterministic case to characterize the approximate symmetric capacity for the symmetric Gaussian bursty IC with feedback in the weak interference regime.},
 author = {I-Hsiang Wang and Changho Suh and Diggavi, S. and Viswanath, P.},
 booktitle = {Information Theory Proceedings (ISIT), 2013 IEEE International Symposium on},
 doi = {10.1109/ISIT.2013.6620180},
 file = {:papers:bursty_ic.pdf},
 issn = {2157-8095},
 month = {July},
 pages = {21-25},
 tags = {conf,WiIntMgmt,IT,WiNet},
 title = {Bursty interference channel with feedback},
 type = {4},
 year = {2013}
}

We explore the benefit of feedback for physical layer interference management in wireless networks without centralized upper layer control mechanisms. Lack of coordination in the upper layer could make the interference experienced in the physical layer bursty. To understand how to harness such burstiness with feedback, we investigate a two-user bursty interference channel (IC), where the presence of interference is governed by a Bernoulli random state. We completely characterize the capacity region of the symmetric two-user linear deterministic bursty IC with feedback. The proposed two-phase scheme exploits feedback either for refining the previous interfered reception or for relaying additional information to the legitimate receiver of the other user. Matching outer bounds are derived by novel techniques that take the effect of delayed state information into account. We also use insights from the deterministic case to characterize the approximate symmetric capacity for the symmetric Gaussian bursty IC with feedback in the weak interference regime.

@inproceedings{6620181,
 abstract = {We investigate how to exploit intermittent feedback for interference management. Focusing on the two-user linear deterministic interference channel, we completely characterize the capacity region. We find that the characterization only depends on the forward channel parameters and the marginal probability distribution of each feedback link. The scheme we propose makes use of block Markov encoding and quantize-map-and-forward at the transmitters, and backward decoding at the receivers. Matching outer bounds are derived based on novel genie-aided techniques. As a consequence, the perfect-feedback capacity can be achieved once the two feedback links are active with large enough probabilities.},
 author = {Karakus, C. and I-Hsiang Wang and Diggavi, S.},
 booktitle = {Information Theory Proceedings (ISIT), 2013 IEEE International Symposium on},
 doi = {10.1109/ISIT.2013.6620181},
 issn = {2157-8095},
 month = {July},
 pages = {26-30},
 tags = {conf,WiIntMgmt,IT,WiNet},
 title = {Interference channel with intermittent feedback},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1305.3265},
 year = {2013}
}

We investigate how to exploit intermittent feedback for interference management. Focusing on the two-user linear deterministic interference channel, we completely characterize the capacity region. We find that the characterization only depends on the forward channel parameters and the marginal probability distribution of each feedback link. The scheme we propose makes use of block Markov encoding and quantize-map-and-forward at the transmitters, and backward decoding at the receivers. Matching outer bounds are derived based on novel genie-aided techniques. As a consequence, the perfect-feedback capacity can be achieved once the two feedback links are active with large enough probabilities.

@inproceedings{6620231,
 abstract = {We study the channel coding problem when errors and uncertainty occur in the encoding process. For simplicity we assume the channel between the encoder and the decoder is perfect. Focusing on linear block codes, we model the encoding uncertainty as erasures on the edges in the factor graph of the encoder generator matrix. We first take a worst-case approach and find the maximum tolerable number of erasures for perfect error correction. Next, we take a probabilistic approach and derive a sufficient condition on the rate of a set of codes, such that decoding error probability vanishes as blocklength tends to infinity. In both scenarios, due to the inherent asymmetry of the problem, we derive the results from first principles, which indicates that robustness to encoding errors requires new properties of codes different from classical properties.},
 author = {Hachem, J. and I-Hsiang Wang and Fragouli, C. and Diggavi, S.},
 booktitle = {Information Theory Proceedings (ISIT), 2013 IEEE International Symposium on},
 doi = {10.1109/ISIT.2013.6620231},
 issn = {2157-8095},
 month = {July},
 pages = {276-280},
 tags = {conf,IT},
 title = {Coding with encoding uncertainty},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1305.3733},
 year = {2013}
}

We study the channel coding problem when errors and uncertainty occur in the encoding process. For simplicity we assume the channel between the encoder and the decoder is perfect. Focusing on linear block codes, we model the encoding uncertainty as erasures on the edges in the factor graph of the encoder generator matrix. We first take a worst-case approach and find the maximum tolerable number of erasures for perfect error correction. Next, we take a probabilistic approach and derive a sufficient condition on the rate of a set of codes, such that decoding error probability vanishes as blocklength tends to infinity. In both scenarios, due to the inherent asymmetry of the problem, we derive the results from first principles, which indicates that robustness to encoding errors requires new properties of codes different from classical properties.

@inproceedings{6620254,
 abstract = {We study opportunistic interference management when there is bursty interference in parallel 2-user linear deterministic interference channels. A degraded message set communication problem is formulated to exploit the burstiness of interference in M subcarriers allocated to each user. We focus on symmetric rate requirements based on the number of interfered subcarriers rather than the exact set of interfered subcarriers. Inner bounds are obtained using erasure coding, signal-scale alignment and Han-Kobayashi coding strategy. Tight outer bounds for a variety of regimes are obtained using the El Gamal-Costa injective interference channel bounds and a sliding window subset entropy inequality [7]. The result demonstrates an application of techniques from multilevel diversity coding to interference channels. We also conjecture outer bounds indicating the sub-optimality of erasure coding across subcarriers in certain regimes.},
 author = {Mishra, S. and I-Hsiang Wang and Diggavi, S.},
 booktitle = {Information Theory Proceedings (ISIT), 2013 IEEE International Symposium on},
 doi = {10.1109/ISIT.2013.6620254},
 issn = {2157-8095},
 month = {July},
 pages = {389-393},
 tags = {conf,WiIntMgmt,IT,WiNet},
 title = {Opportunistic interference management for multicarrier systems},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1305.2985},
 year = {2013}
}

We study opportunistic interference management when there is bursty interference in parallel 2-user linear deterministic interference channels. A degraded message set communication problem is formulated to exploit the burstiness of interference in M subcarriers allocated to each user. We focus on symmetric rate requirements based on the number of interfered subcarriers rather than the exact set of interfered subcarriers. Inner bounds are obtained using erasure coding, signal-scale alignment and Han-Kobayashi coding strategy. Tight outer bounds for a variety of regimes are obtained using the El Gamal-Costa injective interference channel bounds and a sliding window subset entropy inequality [7]. The result demonstrates an application of techniques from multilevel diversity coding to interference channels. We also conjecture outer bounds indicating the sub-optimality of erasure coding across subcarriers in certain regimes.

@inproceedings{6620656,
 abstract = {We study the problem of secure message multicasting over graphs in the presence of a passive (node) adversary who tries to eavesdrop in the network. We show that use of feedback, facilitated through the existence of cycles or undirected edges, enables higher rates than possible in directed acyclic graphs of the same mincut. We demonstrate this using code constructions for canonical combination networks (CCNs). We also provide general outer bounds as well as schemes for node adversaries over CCNs.},
 author = {Mishra, S. and Fragouli, C. and Prabhakaran, V. and Diggavi, S.},
 booktitle = {Information Theory Proceedings (ISIT), 2013 IEEE International Symposium on},
 doi = {10.1109/ISIT.2013.6620656},
 issn = {2157-8095},
 month = {July},
 pages = {2399-2403},
 tags = {conf,WiNetSec,IT},
 title = {Using feedback for secrecy over graphs},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1305.3051},
 year = {2013}
}

We study the problem of secure message multicasting over graphs in the presence of a passive (node) adversary who tries to eavesdrop in the network. We show that use of feedback, facilitated through the existence of cycles or undirected edges, enables higher rates than possible in directed acyclic graphs of the same mincut. We demonstrate this using code constructions for canonical combination networks (CCNs). We also provide general outer bounds as well as schemes for node adversaries over CCNs.

@inproceedings{6620771,
 abstract = {Encoding schemes for broadcasting two nested message sets are studied. We start with a simple class of deterministic broadcast channels for which (variants of) linear superposition coding are optimal in several cases [1], [2]. Such schemes are sub-optimal in general, and we propose a block Markov encoding scheme which achieves (for some deterministic channels) rates not achievable by the previous schemes in [1], [2]. We adapt this block Markov encoding scheme to general broadcast channels, and show that it achieves a rate-region which includes the previously known rate-regions1.},
 author = {Bidokhti, S.S. and Prabhakaran, V.M. and Diggavi, S.N.},
 booktitle = {Information Theory Proceedings (ISIT), 2013 IEEE International Symposium on},
 doi = {10.1109/ISIT.2013.6620771},
 file = {:papers:bme_broadcast.pdf},
 issn = {2157-8095},
 month = {July},
 pages = {2975-2979},
 tags = {conf,IT},
 title = {A block Markov encoding scheme for broadcasting nested message sets},
 type = {4},
 year = {2013}
}

Encoding schemes for broadcasting two nested message sets are studied. We start with a simple class of deterministic broadcast channels for which (variants of) linear superposition coding are optimal in several cases [1], [2]. Such schemes are sub-optimal in general, and we propose a block Markov encoding scheme which achieves (for some deterministic channels) rates not achievable by the previous schemes in [1], [2]. We adapt this block Markov encoding scheme to general broadcast channels, and show that it achieves a rate-region which includes the previously known rate-regions1.

@inproceedings{6691232,
 abstract = {We investigate the problem of secure communication in a simple network with three communicating parties, two distributed sources who communicate over orthogonal channels to one destination node. The cooperation between the sources is restricted to a rate limited common random source they both observe. The communication channels are erasure channels with strictly causal channel state information of the destination available publicly. A passive adversary is present in the system eavesdropping on any one of the channels. We design a linear scheme that ensures secrecy against the eavesdropper. By deriving an outer bound for the problem we prove that the scheme is optimal in certain special cases.},
 author = {Czap, L. and Prabhakaran, V.M. and Diggavi, S. and Fragouli, C.},
 booktitle = {Information Theory Workshop (ITW), 2013 IEEE},
 doi = {10.1109/ITW.2013.6691232},
 file = {:papers:common_randomness.pdf},
 month = {Sept},
 pages = {1-5},
 tags = {conf,WiNetSec,IT},
 title = {Exploiting common randomness: A resource for network secrecy},
 type = {4},
 year = {2013}
}

We investigate the problem of secure communication in a simple network with three communicating parties, two distributed sources who communicate over orthogonal channels to one destination node. The cooperation between the sources is restricted to a rate limited common random source they both observe. The communication channels are erasure channels with strictly causal channel state information of the destination available publicly. A passive adversary is present in the system eavesdropping on any one of the channels. We design a linear scheme that ensures secrecy against the eavesdropper. By deriving an outer bound for the problem we prove that the scheme is optimal in certain special cases.

@inproceedings{6736525,
 abstract = {We consider the two-user Gaussian interference channel with intermittent channel output feedback. We derive an achievable rate region that corresponds to the capacity region of the linear deterministic version of the problem. The result shows that passive and unreliable feedback can be harnessed to provide multiplicative capacity gain in Gaussian interference channels. In contrast to other schemes developed for interference channel with feedback, our achievable scheme makes use of quantize-map-and-forward to relay the information obtained through feedback, performs forward decoding, and does not use structured codes. We find that when the feedback links are active with sufficiently large probabilities, the perfect feedback sum-capacity is achieved to within a constant gap.},
 author = {Karakus, C. and I-Hsiang Wang and Diggavi, S.},
 booktitle = {Communication, Control, and Computing (Allerton), 2013 51st Annual Allerton Conference on},
 doi = {10.1109/Allerton.2013.6736525},
 file = {:papers:ach_gicifb.pdf},
 month = {Oct},
 pages = {203-210},
 tags = {conf,WiIntMgmt,IT,WiNet,WNIF},
 title = {An achievable rate region for Gaussian interference channel with intermittent feedback},
 type = {4},
 year = {2013}
}

We consider the two-user Gaussian interference channel with intermittent channel output feedback. We derive an achievable rate region that corresponds to the capacity region of the linear deterministic version of the problem. The result shows that passive and unreliable feedback can be harnessed to provide multiplicative capacity gain in Gaussian interference channels. In contrast to other schemes developed for interference channel with feedback, our achievable scheme makes use of quantize-map-and-forward to relay the information obtained through feedback, performs forward decoding, and does not use structured codes. We find that when the feedback links are active with sufficiently large probabilities, the perfect feedback sum-capacity is achieved to within a constant gap.

@inproceedings{6736707,
 abstract = {Secure network coding assumes that the underlying network links are lossless, thus it can be applied over lossy networks after channel error correction. Yet it is well known that channel losses, such as packet erasures, can be constructively used for secrecy over a link. We address here the challenge of extending these results for arbitrary networks. We provide achievability schemes over erasure networks with feedback, that outperform the alternative approach of channel error correction followed by secure message transmission separation. We derive outer bounds on the securely achievable rate and as a consequence we show optimality of our proposed scheme in some special cases.},
 author = {Czap, L. and Fragouli, C. and Prabhakaran, V.M. and Diggavi, S.},
 booktitle = {Communication, Control, and Computing (Allerton), 2013 51st Annual Allerton Conference on},
 doi = {10.1109/Allerton.2013.6736707},
 file = {:papers:secure_netcod.pdf},
 month = {Oct},
 pages = {1517-1524},
 tags = {conf,WiNetSec,IT},
 title = {Secure network coding with erasures and feedback},
 type = {4},
 year = {2013}
}

Secure network coding assumes that the underlying network links are lossless, thus it can be applied over lossy networks after channel error correction. Yet it is well known that channel losses, such as packet erasures, can be constructively used for secrecy over a link. We address here the challenge of extending these results for arbitrary networks. We provide achievability schemes over erasure networks with feedback, that outperform the alternative approach of channel error correction followed by secure message transmission separation. We derive outer bounds on the securely achievable rate and as a consequence we show optimality of our proposed scheme in some special cases.

@inproceedings{6810219,
 abstract = {Recent results in wireless full-duplex promise rate gains over the half-duplex counterpart when two nodes exchange messages with each other. However, when multiple full-duplex nodes operate simultaneously, the resulting network has increased internode interference compared to the half-duplex counterpart. The increased internode interference can potentially limit the rate gain achievable due to introduction of full-duplex capability. In this paper, we present new interference management strategies tha handle internode interference for full-duplex enabled network and achieve rate gains over its half-duplex counterpart.},
 author = {Sahai, A and Diggavi, S. and Sabharwal, A},
 booktitle = {Signals, Systems and Computers, 2013 Asilomar Conference on},
 doi = {10.1109/ACSSC.2013.6810219},
 file = {:papers:asilomar_ul_dl_fd.pdf},
 month = {Nov},
 pages = {14-18},
 tags = {conf,WiIntMgmt,IT},
 title = {On uplink/downlink full-duplex networks},
 type = {4},
 year = {2013}
}

Recent results in wireless full-duplex promise rate gains over the half-duplex counterpart when two nodes exchange messages with each other. However, when multiple full-duplex nodes operate simultaneously, the resulting network has increased internode interference compared to the half-duplex counterpart. The increased internode interference can potentially limit the rate gain achievable due to introduction of full-duplex capability. In this paper, we present new interference management strategies tha handle internode interference for full-duplex enabled network and achieve rate gains over its half-duplex counterpart.

@inproceedings{argyraki2013creating,
 abstract = {Current security systems often rely on the adversary's computational limitations. Wireless networks offer the opportunity for a different, complementary kind of security, which relies on the adversary's limited network presence (i.e., that the adversary cannot be located at many different points in the network at the same time). We present a system that leverages this opportunity to enable n wireless nodes to create a shared secret S, in a way that an eavesdropper, Eve, obtains very little information on S. Our system consists of two steps: (1) The nodes transmit packets following a special pattern, such that Eve learns very little about a given fraction of the transmitted packets. This is achieved through a combination of beam forming (from many different sources) and wiretap codes. (2) The nodes participate in a protocol that reshuffles the information known to each node, such that the nodes end up sharing a secret that Eve knows very little about. Our protocol is easily implementable in existing wireless devices and scales well with the number of nodes; these properties are achieved through a combination of public feedback, broadcasting, and network coding. We evaluate our system through a 5-node testbed. We demonstrate that a group of wireless nodes can generate thousands of new shared secret bits per second, with their secrecy being independent of the adversary's computational capabilities.},
 author = {Argyraki, Katerina and Diggavi, Suhas and Duarte, Melissa and Fragouli, Christina and Gatzianas, Marios and Kostopoulos, Panagiotis},
 booktitle = {Proceedings of the 19th annual international conference on Mobile computing \& networking},
 file = {:papers:creating_secrets.pdf},
 month = {Sep},
 organization = {ACM},
 pages = {429--440},
 tags = {conf,WiNetSec,IT},
 title = {Creating secrets out of erasures},
 type = {4},
 year = {2013}
}

Current security systems often rely on the adversary's computational limitations. Wireless networks offer the opportunity for a different, complementary kind of security, which relies on the adversary's limited network presence (i.e., that the adversary cannot be located at many different points in the network at the same time). We present a system that leverages this opportunity to enable n wireless nodes to create a shared secret S, in a way that an eavesdropper, Eve, obtains very little information on S. Our system consists of two steps: (1) The nodes transmit packets following a special pattern, such that Eve learns very little about a given fraction of the transmitted packets. This is achieved through a combination of beam forming (from many different sources) and wiretap codes. (2) The nodes participate in a protocol that reshuffles the information known to each node, such that the nodes end up sharing a secret that Eve knows very little about. Our protocol is easily implementable in existing wireless devices and scales well with the number of nodes; these properties are achieved through a combination of public feedback, broadcasting, and network coding. We evaluate our system through a 5-node testbed. We demonstrate that a group of wireless nodes can generate thousands of new shared secret bits per second, with their secrecy being independent of the adversary's computational capabilities.

@inproceedings{duarte2013quantize,
 abstract = {We present the design and experimental evaluation of a wireless system that exploits relaying in the context of WiFi. We opt for WiFi given its popularity and wide spread use for a number of applications, such as smart homes. Our testbed consists of three nodes, a source, a relay and a destination, that operate using the physical layer procedures of IEEE802.11. We deploy three main competing strategies that have been proposed for relaying, Decode-and-Forward (DF), Amplify-and-Forward (AF) and Quantize-Map-Forward (QMF). QMF is the most recently introduced of the three, and although it was shown in theory to approximately achieve the capacity of arbitrary wireless networks, its performance in practice had not been evaluated. We present in this work experimental results---to the best of our knowledge, the first ones---that compare QMF, AF and DF in a realistic indoor setting. We find that QMF is a competitive scheme to the other two, offering in some cases up to 12\% throughput benefits and up to 60\% improvement in frame error-rates over the next best scheme.},
 author = {Duarte, Melissa and Sengupta, Ayan and Brahma, Siddhartha and Fragouli, Christina and Diggavi, Suhas},
 booktitle = {Proceedings of the fourteenth ACM international symposium on Mobile ad hoc networking and computing},
 file = {:papers:qmf_experimental.pdf},
 month = {Jul},
 organization = {ACM},
 pages = {227--236},
 tags = {conf,WiNet,WiNetInfFlow},
 title = {Quantize-map-forward (QMF) relaying: an experimental study},
 type = {4},
 year = {2013}
}

We present the design and experimental evaluation of a wireless system that exploits relaying in the context of WiFi. We opt for WiFi given its popularity and wide spread use for a number of applications, such as smart homes. Our testbed consists of three nodes, a source, a relay and a destination, that operate using the physical layer procedures of IEEE802.11. We deploy three main competing strategies that have been proposed for relaying, Decode-and-Forward (DF), Amplify-and-Forward (AF) and Quantize-Map-Forward (QMF). QMF is the most recently introduced of the three, and although it was shown in theory to approximately achieve the capacity of arbitrary wireless networks, its performance in practice had not been evaluated. We present in this work experimental results—to the best of our knowledge, the first ones—that compare QMF, AF and DF in a realistic indoor setting. We find that QMF is a competitive scheme to the other two, offering in some cases up to 12% throughput benefits and up to 60% improvement in frame error-rates over the next best scheme.

@inproceedings{sahai2013degrees,
 abstract = {Feasibility of full-duplex opens up the possibility of applying it to cellular networks to operate uplink and downlink simultaneously for multiple users. However, simultaneous operation of uplink and downlink poses a new challenge of intra-cell inter-node interference. In this paper, we identify scenarios where inter-node interference can be managed to provide significant gain in degrees of freedom over the conventional half-duplex cellular design.},
 author = {Sahai, Achaleshwar and Diggavi, Suhas and Sabharwal, Ashutosh},
 booktitle = {Information Theory Workshop (ITW), 2013 IEEE},
 file = {:papers:fd_ul_dl.pdf},
 month = {Sep},
 organization = {IEEE},
 pages = {1--5},
 tags = {conf,WiIntMgmt,IT,WiNet},
 title = {On degrees-of-freedom of full-duplex uplink/downlink channel},
 type = {4},
 year = {2013}
}

Feasibility of full-duplex opens up the possibility of applying it to cellular networks to operate uplink and downlink simultaneously for multiple users. However, simultaneous operation of uplink and downlink poses a new challenge of intra-cell inter-node interference. In this paper, we identify scenarios where inter-node interference can be managed to provide significant gain in degrees of freedom over the conventional half-duplex cellular design.

@inproceedings{wangdiggavi13,
 author = {Wang, I.-H. and Diggavi, S.N.},
 booktitle = {IEEE Information Theory Workshop (ITW) 2013},
 month = {Sep},
 tags = {conf,IT,WiIntMgmt,WiNet},
 title = {Managing bursty interference with feedback},
 type = {4},
 year = {2013}
}

@article{6142074,
 abstract = {Systems that employ network coding for content distribution convey to the receivers linear combinations of the source packets. If we assume randomized network coding, during this process, the network nodes collect random subspaces of the space spanned by the source packets. We establish several fundamental properties of the random subspaces induced in such a system and show that these subspaces implicitly carry topological information about the network and its state that can be passively collected and inferred. We leverage this information toward a number of applications that are interesting in their own right, such as topology inference, bottleneck discovery in peer-to-peer systems, and locating Byzantine attackers. We thus argue that randomized network coding, apart from its better known properties for improving information delivery rate, can additionally facilitate network management and control.},
 author = {Siavoshani, M.J. and Fragouli, C. and Diggavi, S.N.},
 doi = {10.1109/TIT.2012.2184656},
 issn = {0018-9448},
 journal = {Information Theory, IEEE Transactions on},
 month = {May},
 number = {5},
 pages = {2599-2619},
 tags = {journal,NetCod,IT},
 title = {Subspace Properties of Network Coding and Their Applications},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1106.5387},
 volume = {58},
 year = {2012}
}

Systems that employ network coding for content distribution convey to the receivers linear combinations of the source packets. If we assume randomized network coding, during this process, the network nodes collect random subspaces of the space spanned by the source packets. We establish several fundamental properties of the random subspaces induced in such a system and show that these subspaces implicitly carry topological information about the network and its state that can be passively collected and inferred. We leverage this information toward a number of applications that are interesting in their own right, such as topology inference, bottleneck discovery in peer-to-peer systems, and locating Byzantine attackers. We thus argue that randomized network coding, apart from its better known properties for improving information delivery rate, can additionally facilitate network management and control.

@article{6165364,
 abstract = {In this paper, we investigate the maximum achievable sum-rate of the two-user Gaussian interference channel with Gaussian superposition coding and successive decoding. We first examine an approximate deterministic formulation of the problem, and introduce the complementarity conditions that capture the use of Gaussian coding and successive decoding. In the deterministic channel problem, we find the constrained sum-capacity and its achievable schemes with the minimum number of messages, first in symmetric channels, and then in general asymmetric channels. We show that the constrained sum-capacity oscillates as a function of the cross link gain parameters between the information theoretic sum-capacity and the sum-capacity with interference treated as noise. Furthermore, we show that if the number of messages of either of the two users is fewer than the minimum number required to achieve the constrained sum-capacity, the maximum achievable sum-rate drops to that with interference treated as noise. We provide two algorithms to translate the optimal schemes in the deterministic channel model to the Gaussian channel model. We also derive two upper bounds on the maximum achievable sum-rate of the Gaussian Han-Kobayashi schemes, which automatically upper bound the maximum achievable sum-rate using successive decoding of Gaussian codewords. Numerical evaluations show that, similar to the deterministic channel results, the maximum achievable sum-rate with successive decoding in the Gaussian channels oscillates between that with Han-Kobayashi schemes and that with single message schemes.},
 author = {Yue Zhao and Chee-wei Tan and Avestimehr, AS. and Diggavi, S.N. and Pottie, G.J.},
 doi = {10.1109/TIT.2012.2190040},
 file = {:papers:suc_dec_ic.pdf},
 issn = {0018-9448},
 journal = {Information Theory, IEEE Transactions on},
 month = {June},
 number = {6},
 pages = {3798-3820},
 tags = {journal,WiIntMgmt,IT},
 title = {On the Maximum Achievable Sum-Rate With Successive Decoding in Interference Channels},
 type = {2},
 volume = {58},
 year = {2012}
}

In this paper, we investigate the maximum achievable sum-rate of the two-user Gaussian interference channel with Gaussian superposition coding and successive decoding. We first examine an approximate deterministic formulation of the problem, and introduce the complementarity conditions that capture the use of Gaussian coding and successive decoding. In the deterministic channel problem, we find the constrained sum-capacity and its achievable schemes with the minimum number of messages, first in symmetric channels, and then in general asymmetric channels. We show that the constrained sum-capacity oscillates as a function of the cross link gain parameters between the information theoretic sum-capacity and the sum-capacity with interference treated as noise. Furthermore, we show that if the number of messages of either of the two users is fewer than the minimum number required to achieve the constrained sum-capacity, the maximum achievable sum-rate drops to that with interference treated as noise. We provide two algorithms to translate the optimal schemes in the deterministic channel model to the Gaussian channel model. We also derive two upper bounds on the maximum achievable sum-rate of the Gaussian Han-Kobayashi schemes, which automatically upper bound the maximum achievable sum-rate using successive decoding of Gaussian codewords. Numerical evaluations show that, similar to the deterministic channel results, the maximum achievable sum-rate with successive decoding in the Gaussian channels oscillates between that with Han-Kobayashi schemes and that with single message schemes.

@article{kdw_j11,
 abstract = {We study the secret-key capacity in a joint source-channel coding setup-the terminals are connected over a discrete memoryless channel and have access to side information, modelled as a pair of discrete memoryless source sequences. As our main result, we establish the upper and lower bounds on the secret-key capacity. In the lower bound expression, the equivocation terms of the source and channel components are functionally additive even though the coding scheme generates a single secret-key by jointly taking into account the source and channel equivocations. Our bounds coincide, thus establishing the capacity, when the underlying wiretap channel can be decomposed into a set of independent, parallel, and reversely degraded channels. For the case of parallel Gaussian channels and jointly Gaussian sources we show that Gaussian codebooks achieve the secret-key capacity. In addition, when the eavesdropper also observes a correlated side information sequence, we establish the secret-key capacity when both the source and channel of the eavesdropper are a degraded version of the legitimate receiver. We finally also treat the case when a public discussion channel is available, propose a separation based coding scheme, and establish its optimality when the channel output symbols of the legitimate receiver and eavesdropper are conditionally independent given the input.},
 author = {Khisti, A and Diggavi, S.N. and Wornell, Gregory W.},
 doi = {10.1109/TIT.2011.2173629},
 file = {:papers:secret_keygen.pdf},
 issn = {0018-9448},
 journal = {Information Theory, IEEE Transactions on},
 label = {kdw_j11},
 month = {Feb},
 number = {2},
 pages = {652-670},
 tags = {journal,ITsecrecy,KeyGen,IT,WiNetSec,SrcChanSec},
 title = {Secret-Key Generation Using Correlated Sources and Channels},
 type = {2},
 volume = {58},
 year = {2012}
}

We study the secret-key capacity in a joint source-channel coding setup-the terminals are connected over a discrete memoryless channel and have access to side information, modelled as a pair of discrete memoryless source sequences. As our main result, we establish the upper and lower bounds on the secret-key capacity. In the lower bound expression, the equivocation terms of the source and channel components are functionally additive even though the coding scheme generates a single secret-key by jointly taking into account the source and channel equivocations. Our bounds coincide, thus establishing the capacity, when the underlying wiretap channel can be decomposed into a set of independent, parallel, and reversely degraded channels. For the case of parallel Gaussian channels and jointly Gaussian sources we show that Gaussian codebooks achieve the secret-key capacity. In addition, when the eavesdropper also observes a correlated side information sequence, we establish the secret-key capacity when both the source and channel of the eavesdropper are a degraded version of the legitimate receiver. We finally also treat the case when a public discussion channel is available, propose a separation based coding scheme, and establish its optimality when the channel output symbols of the legitimate receiver and eavesdropper are conditionally independent given the input.

@inproceedings{6217871,
 abstract = {In this short review paper we summarize some of our recent results on interactive message secrecy for broadcast erasure channels.},
 author = {Czap, L. and Prabhakaran, V.M. and Diggavi, S. and Fragouli, C.},
 booktitle = {Communications Control and Signal Processing (ISCCSP), 2012 5th International Symposium on},
 doi = {10.1109/ISCCSP.2012.6217871},
 file = {:papers:interactive_secrecy.pdf},
 month = {May},
 pages = {1-4},
 tags = {conf,WiNetSec},
 title = {On interactive secrecy over erasure networks},
 type = {4},
 year = {2012}
}

In this short review paper we summarize some of our recent results on interactive message secrecy for broadcast erasure channels.

@inproceedings{6284217,
 abstract = {In mutiterminal communication systems, signals carrying messages meant for different destinations are often observed together at any given destination receiver. Han and Kobayashi (1981) proposed a receiving strategy which performs a joint unique decoding of messages of interest along with a subset of messages which are not of interest. It is now well-known that this provides an achievable region which is, in general, larger than if the receiver treats all messages not of interest as noise. Nair and El Gamal (2009) and Chong, Motani, Garg, and El Gamal (2008) independently proposed a generalization called indirect or non-unique decoding where the receiver uses the codebook structure of the messages to only uniquely decode its messages of interest. Indirect (non-unique) decoding has since been used in various scenarios. The main result in this paper is to provide an interpretation and a systematic proof technique for why indirect decoding, in all known cases where it has been employed, can be replaced by a particularly designed joint unique decoding strategy, without any penalty from a rate region viewpoint1.},
 author = {Bidokhti, S.S. and Prabhakaran, V.M. and Diggavi, S.N.},
 booktitle = {Information Theory Proceedings (ISIT), 2012 IEEE International Symposium on},
 doi = {10.1109/ISIT.2012.6284217},
 issn = {2157-8095},
 month = {July},
 pages = {398-402},
 tags = {conf,IT},
 title = {Is non-unique decoding necessary?},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1312.4378},
 year = {2012}
}

In mutiterminal communication systems, signals carrying messages meant for different destinations are often observed together at any given destination receiver. Han and Kobayashi (1981) proposed a receiving strategy which performs a joint unique decoding of messages of interest along with a subset of messages which are not of interest. It is now well-known that this provides an achievable region which is, in general, larger than if the receiver treats all messages not of interest as noise. Nair and El Gamal (2009) and Chong, Motani, Garg, and El Gamal (2008) independently proposed a generalization called indirect or non-unique decoding where the receiver uses the codebook structure of the messages to only uniquely decode its messages of interest. Indirect (non-unique) decoding has since been used in various scenarios. The main result in this paper is to provide an interpretation and a systematic proof technique for why indirect decoding, in all known cases where it has been employed, can be replaced by a particularly designed joint unique decoding strategy, without any penalty from a rate region viewpoint1.

@inproceedings{6284220,
 abstract = {The value of relay nodes to enhance the error performance versus rate trade-off in wireless networks has been studied extensively. However, wireless nodes currently are constrained to only transmit or receive at a given frequency, i.e., half-duplex constraint. The diversity-multiplexing tradeoff (DMT) for half-duplex networks are less understood. In the special cases where the DMT is currently known, such as the relay channel and the line network, it is achieved by either dynamic decoding or a quantize-map-forward (QMF) strategy with a fixed half-duplex schedule. The main question we investigate in this paper is whether these two strategies are sufficient to achieve the DMT of half-duplex wireless networks or we need new strategies for general setups. We propose a generalization of the two existing schemes through a dynamic QMF strategy and show that in a parallel relay channel it outperforms both earlier schemes. We also establish the DMT for the relay channel with multiple relays and multiple antennas in some special cases.},
 author = {Ozgur, A and Diggavi, S.},
 booktitle = {Information Theory Proceedings (ISIT), 2012 IEEE International Symposium on},
 doi = {10.1109/ISIT.2012.6284220},
 file = {:papers:dynamic_qmf.pdf},
 issn = {2157-8095},
 month = {July},
 pages = {413-417},
 tags = {conf,WiNet,IT,WiNetInfFlow},
 title = {Dynamic QMF for half-duplex relay networks},
 type = {4},
 year = {2012}
}

The value of relay nodes to enhance the error performance versus rate trade-off in wireless networks has been studied extensively. However, wireless nodes currently are constrained to only transmit or receive at a given frequency, i.e., half-duplex constraint. The diversity-multiplexing tradeoff (DMT) for half-duplex networks are less understood. In the special cases where the DMT is currently known, such as the relay channel and the line network, it is achieved by either dynamic decoding or a quantize-map-forward (QMF) strategy with a fixed half-duplex schedule. The main question we investigate in this paper is whether these two strategies are sufficient to achieve the DMT of half-duplex wireless networks or we need new strategies for general setups. We propose a generalization of the two existing schemes through a dynamic QMF strategy and show that in a parallel relay channel it outperforms both earlier schemes. We also establish the DMT for the relay channel with multiple relays and multiple antennas in some special cases.

@inproceedings{6284224,
 abstract = {Consider a source, Alice, broadcasting private messages to multiple receivers through a broadcast erasure channel; users send back to Alice public feedback that she causally uses to decide the coding strategy for her following transmissions. Recently, the multiple unicast capacity region for this problem has been exactly characterized for a number of special cases; namely the 2-user, 3-user, symmetric K-user, and one-sidedly fair K-user [1], [2]. In this paper, we show that for all the cases where such characterizations exist, we can also optimally characterize the “secure” communication rates, where the message that Alice transmits to each user is information theoretically secure from the other users, even if these collude. We show that a simple, two-phase strategy, where appropriate amounts of secret keys are first generated and then consumed, matches a new outer bound we derive.},
 author = {Czap, L. and Prabhakaran, V.M. and Diggavi, S. and Fragouli, C.},
 booktitle = {Information Theory Proceedings (ISIT), 2012 IEEE International Symposium on},
 doi = {10.1109/ISIT.2012.6284224},
 file = {:papers:bc_private_msg.pdf},
 issn = {2157-8095},
 month = {July},
 pages = {428-432},
 tags = {conf,WiNetSec,IT},
 title = {Broadcasting private messages securely},
 type = {4},
 year = {2012}
}

Consider a source, Alice, broadcasting private messages to multiple receivers through a broadcast erasure channel; users send back to Alice public feedback that she causally uses to decide the coding strategy for her following transmissions. Recently, the multiple unicast capacity region for this problem has been exactly characterized for a number of special cases; namely the 2-user, 3-user, symmetric K-user, and one-sidedly fair K-user [1], [2]. In this paper, we show that for all the cases where such characterizations exist, we can also optimally characterize the “secure” communication rates, where the message that Alice transmits to each user is information theoretically secure from the other users, even if these collude. We show that a simple, two-phase strategy, where appropriate amounts of secret keys are first generated and then consumed, matches a new outer bound we derive.

@inproceedings{6284233,
 abstract = {In this paper we study the two unicast information flow problem over layered Gaussian networks with arbitrary number of nodes and connectivity, under the model of delayed channel state information (CSI) at transmitters and instantaneous CSI at receivers. We show that similar to the case with instantaneous CSI at transmitters (CSIT), the degrees of freedom (DoF) region is strictly larger than the time-sharing DoF region if and only if there is no omniscient node, definition of which only depends on the topology of the network. Moreover, as in the case with instantaneous CSIT, 2/3 DoF per user is always achievable when there is no omniscient node in the network.},
 author = {I-Hsiang Wang and Diggavi, S.},
 booktitle = {Information Theory Proceedings (ISIT), 2012 IEEE International Symposium on},
 doi = {10.1109/ISIT.2012.6284233},
 file = {:papers:dof_2_unicast.pdf},
 issn = {2157-8095},
 month = {July},
 pages = {46-50},
 tags = {conf,WiNet,IT},
 title = {On degrees of freedom of layered two unicast networks with delayed CSIT},
 type = {4},
 year = {2012}
}

In this paper we study the two unicast information flow problem over layered Gaussian networks with arbitrary number of nodes and connectivity, under the model of delayed channel state information (CSI) at transmitters and instantaneous CSI at receivers. We show that similar to the case with instantaneous CSI at transmitters (CSIT), the degrees of freedom (DoF) region is strictly larger than the time-sharing DoF region if and only if there is no omniscient node, definition of which only depends on the topology of the network. Moreover, as in the case with instantaneous CSIT, 2/3 DoF per user is always achievable when there is no omniscient node in the network.

@inproceedings{6426811,
 abstract = {We consider the problem of estimation and control of a linear system when some of the sensors or actuators are attacked by a malicious agent. In our previous work [1] we studied systems with no control inputs and we formulated the estimation problem as a dynamic error correction problem with sparse attack vectors. In this paper we extend our study and look at the role of inputs and control. We first show that it is possible to increase the resilience of the system to attacks by changing the dynamics of the system using state-feedback while having (almost) total freedom in placing the new poles of the system. We then look at the problem of stabilizing a plant using output-feedback despite attacks on sensors, and we show that a principle of separation of estimation and control holds. Finally we look at the effect of attacks on actuators in addition to attacks on sensors: we characterize the resilience of the system with respect to actuator and sensor attacks and we formulate an efficient optimization-based decoder to estimate the state of the system despite attacks on actuators and sensors.},
 author = {Fawzi, H. and Tabuada, P. and Diggavi, S.},
 booktitle = {Decision and Control (CDC), 2012 IEEE 51st Annual Conference on},
 doi = {10.1109/CDC.2012.6426811},
 file = {:papers:security_cdc.pdf},
 issn = {0743-1546},
 month = {Dec},
 pages = {3412-3417},
 tags = {conf,SecureCPS},
 title = {Security for control systems under sensor and actuator attacks},
 type = {4},
 year = {2012}
}

We consider the problem of estimation and control of a linear system when some of the sensors or actuators are attacked by a malicious agent. In our previous work [1] we studied systems with no control inputs and we formulated the estimation problem as a dynamic error correction problem with sparse attack vectors. In this paper we extend our study and look at the role of inputs and control. We first show that it is possible to increase the resilience of the system to attacks by changing the dynamics of the system using state-feedback while having (almost) total freedom in placing the new poles of the system. We then look at the problem of stabilizing a plant using output-feedback despite attacks on sensors, and we show that a principle of separation of estimation and control holds. Finally we look at the effect of attacks on actuators in addition to attacks on sensors: we characterize the resilience of the system with respect to actuator and sensor attacks and we formulate an efficient optimization-based decoder to estimate the state of the system despite attacks on actuators and sensors.

@inproceedings{6483333,
 abstract = {We consider the problem of distributed computation of a target function over a multiple-access channel. If the target and channel functions are matched (i.e., compute the same function), significant performance gains can be obtained by jointly designing the computation and communication tasks. However, in most situations there is mismatch between these two functions. In this work, we analyze the impact of this mismatch on the performance gains achievable with joint computation and communication designs over separation-based designs. We show that for most pairs of target and channel functions there is no such gain, and separation of computation and communication is optimal.},
 author = {Karamchandani, N. and Niesen, U. and Diggavi, S.},
 booktitle = {Communication, Control, and Computing (Allerton), 2012 50th Annual Allerton Conference on},
 doi = {10.1109/Allerton.2012.6483333},
 file = {:papers:comp_mismatch_allerton.pdf},
 month = {Oct},
 pages = {1042-1048},
 tags = {conf,IT},
 title = {Computation over mismatched channels},
 type = {4},
 year = {2012}
}

We consider the problem of distributed computation of a target function over a multiple-access channel. If the target and channel functions are matched (i.e., compute the same function), significant performance gains can be obtained by jointly designing the computation and communication tasks. However, in most situations there is mismatch between these two functions. In this work, we analyze the impact of this mismatch on the performance gains achievable with joint computation and communication designs over separation-based designs. We show that for most pairs of target and channel functions there is no such gain, and separation of computation and communication is optimal.

@inproceedings{6483356,
 abstract = {Optimal precoder design for weighted sum-rate maximization in multiple-input multiple-output interference networks is studied. For this well known non-convex optimization problem, convex approximations based on interference alignment are developed, for both single-beam and multi-beam cases. Precoder design methods that consist of two phases, an interference alignment phase and a post-alignment optimization phase, are proposed. The interference alignment solution is taken as the input to the post-alignment optimization phase. For post-alignment weighted sum-rate maximization, novel iterative distributed algorithms are proposed based on the developed convex approximations. Simulation results show that the proposed algorithms achieve promising weighted sum-rate gains over existing interference alignment algorithms. Interestingly, for the multi-beam case, significant gain is achieved at all SNRs, including the high SNR regime.},
 author = {Yue Zhao and Diggavi, S.N. and Goldsmith, A and Poor, H.V.},
 booktitle = {Communication, Control, and Computing (Allerton), 2012 50th Annual Allerton Conference on},
 doi = {10.1109/Allerton.2012.6483356},
 file = {:papers:convex_opt_precoder.pdf},
 month = {Oct},
 pages = {1213-1219},
 tags = {conf,WiNet,IT},
 title = {Convex optimization for precoder design in MIMO interference networks},
 type = {4},
 year = {2012}
}

Optimal precoder design for weighted sum-rate maximization in multiple-input multiple-output interference networks is studied. For this well known non-convex optimization problem, convex approximations based on interference alignment are developed, for both single-beam and multi-beam cases. Precoder design methods that consist of two phases, an interference alignment phase and a post-alignment optimization phase, are proposed. The interference alignment solution is taken as the input to the post-alignment optimization phase. For post-alignment weighted sum-rate maximization, novel iterative distributed algorithms are proposed based on the developed convex approximations. Simulation results show that the proposed algorithms achieve promising weighted sum-rate gains over existing interference alignment algorithms. Interestingly, for the multi-beam case, significant gain is achieved at all SNRs, including the high SNR regime.

@inproceedings{safaka2012creating,
 abstract = {Current security systems typically rely on the adversary's computational limitations (e.g., the fact that it cannot invert a hash function or perform large-integer factorization). Wireless networks offer the opportunity for a different, complementary kind of security, which relies not on the adversary's computational limitations, but on its limited network presence (i.e., that the adversary cannot be located at many different points in the network at the same time). We take a first step toward designing and building a wireless security system that leverages this opportunity: We consider the problem where a group of n nodes, connected to the same broadcast wireless network, want to agree on a shared secret (e.g., an encryption key), in the presence of an adversary Eve who tries to listen in and steal the secret. We propose a secret-agreement protocol, where the n nodes of the group keep exchanging bits until they have all agreed on a bit sequence that Eve cannot reconstruct (with very high probability). We provide experimental evidence---to the best of our knowledge, the first one---that a group of wireless nodes can generate thousands of new shared secret bits per second, with their secrecy being independent of the adversary's computational capabilities.},
 author = {Safaka, Iris and Fragouli, Christina and Argyraki, Katerina and Diggavi, Suhas},
 booktitle = {Proceedings of the 11th ACM Workshop on Hot Topics in Networks},
 file = {:papers:secrets_thin_air.pdf},
 month = {Oct},
 organization = {ACM},
 pages = {73--78},
 tags = {conf,WiNetSec,IT},
 title = {Creating shared secrets out of thin air},
 type = {4},
 year = {2012}
}

Current security systems typically rely on the adversary's computational limitations (e.g., the fact that it cannot invert a hash function or perform large-integer factorization). Wireless networks offer the opportunity for a different, complementary kind of security, which relies not on the adversary's computational limitations, but on its limited network presence (i.e., that the adversary cannot be located at many different points in the network at the same time). We take a first step toward designing and building a wireless security system that leverages this opportunity: We consider the problem where a group of n nodes, connected to the same broadcast wireless network, want to agree on a shared secret (e.g., an encryption key), in the presence of an adversary Eve who tries to listen in and steal the secret. We propose a secret-agreement protocol, where the n nodes of the group keep exchanging bits until they have all agreed on a bit sequence that Eve cannot reconstruct (with very high probability). We provide experimental evidence—to the best of our knowledge, the first one—that a group of wireless nodes can generate thousands of new shared secret bits per second, with their secrecy being independent of the adversary's computational capabilities.

@article{ADTj10,
 abstract = {This paper approximately resolves the information flow over wireless (Gaussian) relay networks to within a constant number of bits. It introduced the linear deterministic model and established a max-flow min-cut result for linear deterministic networks. The insights from the deterministic approach led to the approximate max-flow min-cut characterization for noisy Gaussian networks, through the introduction of a new relaying
strategy called {\em quantize-map-forward} (QMF).},
 author = {S. Avestimehr and S N. Diggavi and D N C. Tse},
 file = {:papers:wirelessnetinfflow.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {adtj10},
 month = {April},
 note = {},
 number = {4},
 pages = {},
 tags = {journal,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow},
 title = {Wireless network information flow: a deterministic approach},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0906.5394},
 volume = {57},
 year = {2011}
}

This paper approximately resolves the information flow over wireless (Gaussian) relay networks to within a constant number of bits. It introduced the linear deterministic model and established a max-flow min-cut result for linear deterministic networks. The insights from the deterministic approach led to the approximate max-flow min-cut characterization for noisy Gaussian networks, through the introduction of a new relaying strategy called \em quantize-map-forward (QMF).

@article{DDj11,
 abstract = {We show that the diversity-multiplexing tradeoff in fading inter-symbol interference (ISI) channels with a single degree of freedom (MISO/SIMO) is
successively refinable. This demonstrates that ideal opportunistic codes can be designed
asymptotically in SNR.},
 author = {S. Dusad and S N. Diggavi},
 file = {:papers:ddsucrefisi.pdf},
 journal = {IEEE Transactions on Communications},
 month = {December},
 pages = {1820-1829},
 tags = {journal,DivEmb,SucRefDiv,STcodes,OppComm,DegMsgSet},
 title = {On successive refinement of diversity for fading ISI channels},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0912.4556},
 volume = {59},
 year = {2011}
}

We show that the diversity-multiplexing tradeoff in fading inter-symbol interference (ISI) channels with a single degree of freedom (MISO/SIMO) is successively refinable. This demonstrates that ideal opportunistic codes can be designed asymptotically in SNR.

@article{JMFDj11,
 abstract = {We propose a channel model for the non-coherent network coding introduced by Koetter and Kschischang.  The model captures the essence of such a network operation, and we use it to calculate the capacity of non-coherent network codes as a function of network parameters. We prove that use of subspace coding is optimal, and show that, in
some cases, the capacity-achieving distribution uses subspaces of several dimensions, where the employed dimensions depend on the packet length.},
 author = {M. Jafarisiavoshani and S. Mohajer and C. Fragouli and S N. Diggavi},
 file = {:papers:jmfdnoncoherent10.pdf},
 journal = {IEEE Transactions on Information Theory},
 month = {February},
 number = {2},
 pages = {},
 tags = {journal,NonCohComm,NetCod,Subsp,IT},
 title = {Capacity of noncoherent multisource network coding},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1001.1658},
 volume = {57},
 year = {2011}
}

We propose a channel model for the non-coherent network coding introduced by Koetter and Kschischang. The model captures the essence of such a network operation, and we use it to calculate the capacity of non-coherent network codes as a function of network parameters. We prove that use of subspace coding is optimal, and show that, in some cases, the capacity-achieving distribution uses subspaces of several dimensions, where the employed dimensions depend on the packet length.

@article{KDWj10,
 author = {A. Khisti and S N. Diggavi and G W. Wornell},
 journal = {IEEE Transactions  on Information Forensics and Security},
 month = {Sep},
 note = {},
 pages = {672-681},
 tags = {journal,ITsecrecy,KeyGen,IT,WiNetSec},
 title = {Secret-key agreement over wiretap channels with random state parameters},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1009.3052},
 volume = {6},
 year = {2011}
}

@article{MDFTj10,
 abstract = {In this paper we study  two-stage Gaussian 
relay-interference networks where there are weak cross-links, causing the networks to behave like a chain
of Z Gaussian channels. Our main result is an approximate characterization of the capacity region for  such networks. We propose a new interference management scheme, termed interference
neutralization, which is implemented using structured lattice codes. This scheme allows for over-the-air
interference removal, without the transmitters having complete access the interfering signals.},
 author = {S. Mohajer and S N. Diggavi and C. Fragouli and D. Tse},
 file = {:papers:relayinterference_revised.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {mdft_js10},
 month = {May},
 note = {},
 number = {5},
 pages = {2837-2864},
 tags = {journal,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow,RelayInt,LatticeChan},
 title = {Approximate capacity region for a class of relay-interference networks},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1005.0404},
 volume = {57},
 year = {2011}
}

In this paper we study two-stage Gaussian relay-interference networks where there are weak cross-links, causing the networks to behave like a chain of Z Gaussian channels. Our main result is an approximate characterization of the capacity region for such networks. We propose a new interference management scheme, termed interference neutralization, which is implemented using structured lattice codes. This scheme allows for over-the-air interference removal, without the transmitters having complete access the interfering signals.

@article{tcds_js10,
 abstract = {We provide a complete characterization of the achievable distortion region for the problem of sending a bivariate Gaussian source over bandwidth-matched Gaussian broadcast channels, where each receiver is interested in only one component of the source. This setting naturally generalizes the simple single Gaussian source bandwidth-matched broadcast problem for which the uncoded scheme is known to be optimal. We show that a hybrid scheme can achieve the optimum for the bivariate case, but neither an uncoded scheme alone nor a separation-based scheme alone is sufficient. We further show that in this joint source channel coding setting, the Gaussian scenario is the worst scenario among the sources and channel noises with the same covariances.},
 author = {Tian, C. and Diggavi, S. and Shamai, S.},
 doi = {10.1109/TIT.2011.2165801},
 issn = {0018-9448},
 journal = {Information Theory, IEEE Transactions on},
 label = {tcds_js10},
 month = {Oct},
 number = {10},
 pages = {6419-6427},
 tags = {journal,NDC,WiNet,IT},
 title = {The Achievable Distortion Region of Sending a Bivariate Gaussian Source on the Gaussian Broadcast Channel},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/1006.0644},
 volume = {57},
 year = {2011}
}

We provide a complete characterization of the achievable distortion region for the problem of sending a bivariate Gaussian source over bandwidth-matched Gaussian broadcast channels, where each receiver is interested in only one component of the source. This setting naturally generalizes the simple single Gaussian source bandwidth-matched broadcast problem for which the uncoded scheme is known to be optimal. We show that a hybrid scheme can achieve the optimum for the bivariate case, but neither an uncoded scheme alone nor a separation-based scheme alone is sufficient. We further show that in this joint source channel coding setting, the Gaussian scenario is the worst scenario among the sources and channel noises with the same covariances.

@article{TDSj10a,
 abstract = {We show that source-channel separation is approximately optimal for transmitting
a Gaussian source over a bandwidth mismatched Gaussian channel, for quadratic distortion
measures. Furthermore, we show that the results can be extended to
general broadcast channels, and the performance of the source-channel
separation based approach is also within the same
constant multiplicative factors of the optimum.},
 author = {C. Tian and  S N. Diggavi and S. Shamai(Shitz)},
 file = {:papers:gaussianscbc_final.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {tds_j11a},
 month = {January},
 note = {},
 number = {1},
 pages = {124-136},
 tags = {journal,SrcChan,ApproxIT,NDC,IT},
 title = {Approximate Characterizations for the Gaussian Broadcasting Distortion Region},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0906.3183},
 volume = {57},
 year = {2011}
}

We show that source-channel separation is approximately optimal for transmitting a Gaussian source over a bandwidth mismatched Gaussian channel, for quadratic distortion measures. Furthermore, we show that the results can be extended to general broadcast channels, and the performance of the source-channel separation based approach is also within the same constant multiplicative factors of the optimum.

@inproceedings{6033790,
 abstract = {In this paper, we investigate the sum-capacity of the two-user Gaussian interference channel with Gaussian superposition coding and successive decoding. We first examine an approximate deterministic formulation of the problem, and introduce the complementarity conditions that capture the use of Gaussian coding and successive decoding. In the deterministic channel problem, we show that the constrained sum-capacity oscillates as a function of the cross link gain parameters between the information theoretic sum-capacity and the sum-capacity with interference treated as noise. Furthermore, we show that if the number of messages of either user is fewer than the minimum number required to achieve the constrained sum-capacity, the maximum achievable sum-rate drops to that with interference treated as noise. We translate the optimal schemes in the deterministic channel model to the Gaussian channel model, and also derive two upper bounds on the constrained sum-capacity. Numerical evaluations show that the constrained sum-capacity in the Gaussian channels oscillates between the sum-capacity with Gaussian Han-Kobayashi schemes and that with single message schemes.},
 author = {Yue Zhao and Chee-wei Tan and Avestimehr, AS. and Diggavi, S.N. and Pottie, G.J.},
 booktitle = {Information Theory Proceedings (ISIT), 2011 IEEE International Symposium on},
 doi = {10.1109/ISIT.2011.6033790},
 file = {:papers:suc_dec_ic_isit.pdf},
 issn = {2157-8095},
 month = {July},
 pages = {1494-1498},
 tags = {conf,WiIntMgmt,IT,WiNet},
 title = {On the sum-capacity with successive decoding in interference channels},
 type = {4},
 year = {2011}
}

In this paper, we investigate the sum-capacity of the two-user Gaussian interference channel with Gaussian superposition coding and successive decoding. We first examine an approximate deterministic formulation of the problem, and introduce the complementarity conditions that capture the use of Gaussian coding and successive decoding. In the deterministic channel problem, we show that the constrained sum-capacity oscillates as a function of the cross link gain parameters between the information theoretic sum-capacity and the sum-capacity with interference treated as noise. Furthermore, we show that if the number of messages of either user is fewer than the minimum number required to achieve the constrained sum-capacity, the maximum achievable sum-rate drops to that with interference treated as noise. We translate the optimal schemes in the deterministic channel model to the Gaussian channel model, and also derive two upper bounds on the constrained sum-capacity. Numerical evaluations show that the constrained sum-capacity in the Gaussian channels oscillates between the sum-capacity with Gaussian Han-Kobayashi schemes and that with single message schemes.

@inproceedings{6033895,
 abstract = {We consider a group of m trusted nodes that aim to create a shared secret key K, using a state-dependent wireless broadcast channel that exists from one of the honest nodes to the rest of the nodes including a passive eavesdropper Eve. All of the trusted nodes can also discuss over a cost-free and unlimited rate public channel which is also observed by Eve. For this setup, we develop an information-theoretically secure secret key agreement protocol. We show the optimality of this protocol for linear deterministic wireless broadcast channels as well as in the high-SNR regime for wireless channels with large dynamic range over channel states.},
 author = {Siavoshani, M.J. and Mishra, S. and Diggavi, S.N. and Fragouli, C.},
 booktitle = {Information Theory Proceedings (ISIT), 2011 IEEE International Symposium on},
 doi = {10.1109/ISIT.2011.6033895},
 file = {:papers:group_secret_key.pdf},
 issn = {2157-8095},
 month = {July},
 pages = {1960-1964},
 tags = {conf,WiNetSec,IT},
 title = {Group secret key agreement over state-dependent wireless broadcast channels},
 type = {4},
 year = {2011}
}

We consider a group of m trusted nodes that aim to create a shared secret key K, using a state-dependent wireless broadcast channel that exists from one of the honest nodes to the rest of the nodes including a passive eavesdropper Eve. All of the trusted nodes can also discuss over a cost-free and unlimited rate public channel which is also observed by Eve. For this setup, we develop an information-theoretically secure secret key agreement protocol. We show the optimality of this protocol for linear deterministic wireless broadcast channels as well as in the high-SNR regime for wireless channels with large dynamic range over channel states.

@inproceedings{6034040,
 abstract = {We consider the Gaussian “diamond” or parallel relay network, in which a source node transmits a message to a destination node with the help of N relays. Even for the symmetric setting, in which the channel gains to the relays are identical and the channel gains from the relays are identical, the capacity of this channel is unknown in general. The best known capacity approximation is up to an additive gap of order N bits and up to a multiplicative gap of order N2, with both gaps independent of the channel gains. In this paper, we approximate the capacity of the symmetric Gaussian N-relay diamond network up to an additive gap of 1.8 bits and up to a multiplicative gap of a factor 14. Both gaps are independent of the channel gains, and, unlike the best previously known result, are also independent of the number of relays N in the network. Achievability is based on bursty amplify-and-forward, showing that this simple scheme is uniformly approximately optimal, both in the low-rate as well as high-rate regimes. The upper bound on capacity is based on a careful evaluation of the cut-set bound.},
 author = {Niesen, U. and Diggavi, S.},
 booktitle = {Information Theory Proceedings (ISIT), 2011 IEEE International Symposium on},
 doi = {10.1109/ISIT.2011.6034040},
 file = {:papers:nrelay_diamond_isit.pdf},
 issn = {2157-8095},
 month = {July},
 pages = {259-263},
 tags = {conf,approxIT,IT,WiNet},
 title = {The approximate capacity of the Gaussian N-relay diamond network},
 type = {4},
 year = {2011}
}

We consider the Gaussian “diamond” or parallel relay network, in which a source node transmits a message to a destination node with the help of N relays. Even for the symmetric setting, in which the channel gains to the relays are identical and the channel gains from the relays are identical, the capacity of this channel is unknown in general. The best known capacity approximation is up to an additive gap of order N bits and up to a multiplicative gap of order N2, with both gaps independent of the channel gains. In this paper, we approximate the capacity of the symmetric Gaussian N-relay diamond network up to an additive gap of 1.8 bits and up to a multiplicative gap of a factor 14. Both gaps are independent of the channel gains, and, unlike the best previously known result, are also independent of the number of relays N in the network. Achievability is based on bursty amplify-and-forward, showing that this simple scheme is uniformly approximately optimal, both in the low-rate as well as high-rate regimes. The upper bound on capacity is based on a careful evaluation of the cut-set bound.

@inproceedings{6089363,
 abstract = {We present a structured Quantize-Map-and-Forward (QMF) scheme for cooperative communication over wireless networks, that employs LDPC ensembles for the node operations and message-passing algorithms for decoding. We demonstrate through extensive simulation results over the full-duplex parallel relay network, that our scheme, with no transmit channel state information, offers a robust performance over fading channels and achieves the full diversity order of our network at moderate SNRs.},
 author = {Sengupta, A and Brahma, S. and Ozgur, A and Fragouli, C. and Diggavi, S.},
 booktitle = {Information Theory Workshop (ITW), 2011 IEEE},
 doi = {10.1109/ITW.2011.6089363},
 file = {:papers:graph_qmf.pdf},
 month = {Oct},
 pages = {140-144},
 tags = {conf,IT,WiNet,WiNetInfFlow},
 title = {Graph-based codes for Quantize-Map-and-Forward relaying},
 type = {4},
 year = {2011}
}

We present a structured Quantize-Map-and-Forward (QMF) scheme for cooperative communication over wireless networks, that employs LDPC ensembles for the node operations and message-passing algorithms for decoding. We demonstrate through extensive simulation results over the full-duplex parallel relay network, that our scheme, with no transmit channel state information, offers a robust performance over fading channels and achieves the full diversity order of our network at moderate SNRs.

@inproceedings{6089579,
 abstract = {We characterize the secret message capacity of a wiretapped erasure channel where causal channel state information of the honest nodes is publicly available. In doing so, we establish an intimate connection between message secrecy and secret key generation for the same channel setup. We propose a linear coding scheme that has polynomial encoding/decoding complexity, and prove a converse that shows the optimality of our scheme. Our work also demonstrates the value of causal public feedback, which has previously been shown for the secret key generation problem.},
 author = {Czap, L. and Prabhakaran, V.M. and Fragouli, C. and Diggavi, S.},
 booktitle = {Information Theory Workshop (ITW), 2011 IEEE},
 file = {:papers:secret_erasure_bc.pdf},
 month = {Oct},
 pages = {65-69},
 tags = {conf,WiNetSec,IT},
 title = {Secret message capacity of erasure broadcast channels with feedback},
 type = {4},
 year = {2011}
}

We characterize the secret message capacity of a wiretapped erasure channel where causal channel state information of the honest nodes is publicly available. In doing so, we establish an intimate connection between message secrecy and secret key generation for the same channel setup. We propose a linear coding scheme that has polynomial encoding/decoding complexity, and prove a converse that shows the optimality of our scheme. Our work also demonstrates the value of causal public feedback, which has previously been shown for the secret key generation problem.

@inproceedings{6120187,
 abstract = {We consider the problem of state-estimation of a linear dynamical system when some of the sensor measurements are corrupted by an adversarial attacker. The errors injected by the attacker in the sensor measurements can be arbitrary and are not assumed to follow a specific model (in particular they can be of arbitrary magnitude). We first characterize the number of attacked sensors that can be tolerated so that the state of the system can still be correctly recovered by any decoding algorithm. We then propose a specific computationally feasible decoding algorithm and we give a characterization of the number of errors this decoder can correct. For this we use ideas from compressed sensing and error correction over the reals and we exploit the dynamical nature of the problem. We show using numerical simulations that this decoder performs very well in practice and allows to correct a large number of errors.},
 author = {Fawzi, H. and Tabuada, P. and Diggavi, S.},
 booktitle = {Communication, Control, and Computing (Allerton), 2011 49th Annual Allerton Conference on},
 doi = {10.1109/Allerton.2011.6120187},
 file = {:papers:secure_state_estimate.pdf},
 month = {Sept},
 pages = {337-344},
 tags = {conf,SecureCPS},
 title = {Secure state-estimation for dynamical systems under active adversaries},
 type = {4},
 year = {2011}
}

We consider the problem of state-estimation of a linear dynamical system when some of the sensor measurements are corrupted by an adversarial attacker. The errors injected by the attacker in the sensor measurements can be arbitrary and are not assumed to follow a specific model (in particular they can be of arbitrary magnitude). We first characterize the number of attacked sensors that can be tolerated so that the state of the system can still be correctly recovered by any decoding algorithm. We then propose a specific computationally feasible decoding algorithm and we give a characterization of the number of errors this decoder can correct. For this we use ideas from compressed sensing and error correction over the reals and we exploit the dynamical nature of the problem. We show using numerical simulations that this decoder performs very well in practice and allows to correct a large number of errors.

@inproceedings{6120197,
 abstract = {Information-theoretic secrecy is studied for a parallel relay (diamond) network, in which a transmitter wishes to communicate to a receiver through two relay nodes. While there is no direct link between the transmitter and receiver and all flow of information has to be transmitted through the relays, the message has to be kept secret from each of them. The exact secrecy capacity is characterized for the network under the linear deterministic model. The problem is then studied when each terminal is equipped with multiple antennas, and the channels are parallel Gaussian links. Lower and upper bounds for the secrecy capacity are derived, and the gap is bounded by a constant independent of the channel parameters and SNR. This results in an approximate characterization for the secrecy capacity of the parallel Gaussian diamond network.},
 author = {Mohajer, S. and Diggavi, S.N. and Poor, H.V. and Shamai, S.},
 booktitle = {Communication, Control, and Computing (Allerton), 2011 49th Annual Allerton Conference on},
 doi = {10.1109/Allerton.2011.6120197},
 file = {:papers:parallel_wiretap.pdf},
 month = {Sept},
 pages = {418-425},
 tags = {conf,IT,WiNetSec,WiNet},
 title = {On the parallel relay wire-tap network},
 type = {4},
 year = {2011}
}

Information-theoretic secrecy is studied for a parallel relay (diamond) network, in which a transmitter wishes to communicate to a receiver through two relay nodes. While there is no direct link between the transmitter and receiver and all flow of information has to be transmitted through the relays, the message has to be kept secret from each of them. The exact secrecy capacity is characterized for the network under the linear deterministic model. The problem is then studied when each terminal is equipped with multiple antennas, and the channels are parallel Gaussian links. Lower and upper bounds for the secrecy capacity are derived, and the gap is bounded by a constant independent of the channel parameters and SNR. This results in an approximate characterization for the secrecy capacity of the parallel Gaussian diamond network.

@inproceedings{6162578,
 abstract = {We study secret-key agreement with public discussion over a flat-fading wiretap channel model. The fading gains are correlated across the receivers and sampled independently at each time. Perfect receiver channel state information (CSI) is assumed, whereas a noisy CSI of the main channel is also available to the transmitter. We propose lower and upper bounds on the capacity. Our lower bound is achieved by a coding scheme that involves a separate binning of the receiver CSI sequence and its channel output sequence. In general it improves upon the joint-binning schemes considered in earlier works. Our upper and lower bounds coincide, establishing the capacity, when either the transmitter has no CSI or when the channel gains of the legitimate receiver and the eavesdropper are statistically independent.},
 author = {Khisti, A and Diggavi, S.N.},
 booktitle = {GLOBECOM Workshops (GC Wkshps), 2011 IEEE},
 doi = {10.1109/GLOCOMW.2011.6162578},
 file = {:papers:remark_secret_keygen.pdf},
 month = {Dec},
 pages = {864-868},
 tags = {conf,WiNetSec,IT},
 title = {A remark on secret-key generation over correlated fading channels},
 type = {4},
 year = {2011}
}

We study secret-key agreement with public discussion over a flat-fading wiretap channel model. The fading gains are correlated across the receivers and sampled independently at each time. Perfect receiver channel state information (CSI) is assumed, whereas a noisy CSI of the main channel is also available to the transmitter. We propose lower and upper bounds on the capacity. Our lower bound is achieved by a coding scheme that involves a separate binning of the receiver CSI sequence and its channel output sequence. In general it improves upon the joint-binning schemes considered in earlier works. Our upper and lower bounds coincide, establishing the capacity, when either the transmitter has no CSI or when the channel gains of the legitimate receiver and the eavesdropper are statistically independent.

@inproceedings{tschopp2011randomized,
 author = {Tschopp, Dominique and Diggavi, Suhas and Delgosha, Payam and Mohajer, Soheil},
 booktitle = {Advances in Neural Information Processing Systems},
 file = {:papers:rand_alg_search.pdf},
 month = {Dec},
 pages = {2231--2239},
 tags = {conf,CSalgo,DynamWiNet},
 title = {Randomized algorithms for comparison-based search},
 type = {4},
 year = {2011}
}

@article{KJFADj10,
 abstract = {In a significant class of sensor-network applications,
the identities of the reporting sensors constitute the bulk of
the communicated data, whereas the message itself can be as
small as a single bit. or to track incremental environmental changes at fixed
locations. In such scenarios, the traditional network-protocol
paradigm of separately specifying the source identity and the
message in distinct fields leads to inefficient communication.
This work addresses the question of how communication
should happen in such identity-aware sensor networks.},
 author = {L. Keller and M. Jafarisiavoshani and  C. Fragouli and A. Argyraki and S N. Diggavi},
 file = {:papers:jmfdnoncoherent10.pdf},
 journal = {IEEE Journal on Selected Areas in Communications},
 month = {September},
 note = {},
 number = {7},
 pages = {1083--1093},
 tags = {journal,SensNet,IDmsg,NetCod},
 title = {Joint identity-message coding for sensor networks},
 type = {2},
 volume = {28},
 year = {2010}
}

In a significant class of sensor-network applications, the identities of the reporting sensors constitute the bulk of the communicated data, whereas the message itself can be as small as a single bit. or to track incremental environmental changes at fixed locations. In such scenarios, the traditional network-protocol paradigm of separately specifying the source identity and the message in distinct fields leads to inefficient communication. This work addresses the question of how communication should happen in such identity-aware sensor networks.

@article{LTDCCj10,
 abstract = {We study end-to-end network utility maximization (NUM) approach to allocate
resources to multiple flows, each encoding multiple description data compression streams,
by taking into account the end-to-end distortion measures.},
 author = {Y. Li and C. Tian and S N. Diggavi and M. Chiang and  A. R. Calderbank},
 file = {:papers:competingmd_final.pdf},
 journal = {IEEE Transactions on Communications},
 month = {MAY},
 note = {},
 number = {5},
 pages = {1493--1504},
 tags = {journal,MDnum,NDC,MDC},
 title = {Network Resource Allocation for Competing Multiple-Description Transmissions},
 type = {2},
 volume = {58},
 year = {2010}
}

We study end-to-end network utility maximization (NUM) approach to allocate resources to multiple flows, each encoding multiple description data compression streams, by taking into account the end-to-end distortion measures.

@article{MTDj10a,
 abstract = {This paper introduces the asymmetric multi-level lossless data compression
problem and gives a complete characterization for the three-description version. It then
uses insights from this characterization to provide an approximate characterization for
the 3-description asymmetric multiple description problem.},
 author = {S. Mohajer and C. Tian and S N. Diggavi},
 file = {:papers:asymmd.pdf},
 journal = {IEEE Transactions on Information Theory},
 month = {September},
 note = {},
 pages = {4367-4387},
 tags = {journal,AsymMDapprox,ApproxIT,NDC,MDapprox,IT},
 title = {Asymmetric Multilevel Diversity Coding and  Gaussian Multiple Descriptions},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0911.2346},
 volume = {9},
 year = {2010}
}

This paper introduces the asymmetric multi-level lossless data compression problem and gives a complete characterization for the three-description version. It then uses insights from this characterization to provide an approximate characterization for the 3-description asymmetric multiple description problem.

@inproceedings{5706949,
 abstract = {Hierarchical cooperation is a communication strategy introduced recently by Özgür et al., where it was shown to significantly outperform standard multi-hop communication in some wireless scenarios. To achieve this gain over multi-hop communication, full availability of channel state information at all the nodes in the network is required. In this paper, we investigate the impact of absence of channel state information on the performance of hierarchical cooperation. If the product of coherence time and coherence bandwidth of the wireless channel is on the order of the number of nodes in the network, we show that non-coherent hierarchical cooperation achieves the same order rate as the coherent version. If the product is smaller than the number of nodes, we show that non-coherent hierarchical cooperation can still yield sizable gains over multihop communication, but may not achieve the same order rate as the coherent version.},
 author = {Niesen, U. and Diggavi, S.},
 booktitle = {Communication, Control, and Computing (Allerton), 2010 48th Annual Allerton Conference on},
 doi = {10.1109/ALLERTON.2010.5706949},
 month = {Sept},
 pages = {507-513},
 tags = {conf,WiNet,IT},
 title = {Non-coherent hierarchical cooperation},
 type = {4},
 year = {2010}
}

Hierarchical cooperation is a communication strategy introduced recently by Özgür et al., where it was shown to significantly outperform standard multi-hop communication in some wireless scenarios. To achieve this gain over multi-hop communication, full availability of channel state information at all the nodes in the network is required. In this paper, we investigate the impact of absence of channel state information on the performance of hierarchical cooperation. If the product of coherence time and coherence bandwidth of the wireless channel is on the order of the number of nodes in the network, we show that non-coherent hierarchical cooperation achieves the same order rate as the coherent version. If the product is smaller than the number of nodes, we show that non-coherent hierarchical cooperation can still yield sizable gains over multihop communication, but may not achieve the same order rate as the coherent version.

@inproceedings{5757657,
 abstract = {We consider a group of m trusted nodes that aim to create a shared secret key K over a wireless channel in the presence an eavesdropper Eve. We assume an erasure broadcast channel from one of the honest nodes to the rest of them including Eve. All of the trusted nodes can also discuss over a cost-free public channel which is observed by Eve. For this setup we characterize the secret key generation capacity and propose an achievability scheme that is computationally efficient and employs techniques from network coding. Surprisingly, whether we have m = 2 nodes, or an arbitrary number m of nodes, we can establish a shared secret key among them at the same rate, independently of m.},
 author = {Siavoshani, M.J. and Fragouli, C. and Diggavi, S. and Pulleti, U. and Argyraki, K.},
 booktitle = {Signals, Systems and Computers (ASILOMAR), 2010 Conference Record of the Forty Fourth Asilomar Conference on},
 doi = {10.1109/ACSSC.2010.5757657},
 issn = {1058-6393},
 month = {Nov},
 pages = {719-723},
 tags = {conf,IT,WiNetSec},
 title = {Group secret key generation over broadcast erasure channels},
 type = {4},
 year = {2010}
}

We consider a group of m trusted nodes that aim to create a shared secret key K over a wireless channel in the presence an eavesdropper Eve. We assume an erasure broadcast channel from one of the honest nodes to the rest of them including Eve. All of the trusted nodes can also discuss over a cost-free public channel which is observed by Eve. For this setup we characterize the secret key generation capacity and propose an achievability scheme that is computationally efficient and employs techniques from network coding. Surprisingly, whether we have m = 2 nodes, or an arbitrary number m of nodes, we can establish a shared secret key among them at the same rate, independently of m.

@inproceedings{JFDPA10,
 author = {M. Jafarisiavoshani and C. Fragouli and S N. Diggavi and U. Pulleti and K. Argyraki},
 booktitle = {Proc. Asilomar Conference on Signals, Systems and  Computers, Pacific Grove, CA},
 month = {November},
 note = {},
 pages = {},
 tags = {conf},
 title = {Group secret key generation over broadcast erasure channels (invited talk)},
 type = {4},
 year = {2010}
}

@inproceedings{MDc10a,
 abstract = {We study the problem where there is a single adversarial node which injects
signals to disrupt this communication in a parallel relay network. Like the source, it can
only influence the destination through the relays. We develop
an approximate characterization of the reliable transmission
rate in the presence of such an adversary. A deterministic version
of the same problem is solved exactly, yielding insights which are
used in the approximate characterization.},
 author = {S. Mohajer and S N. Diggavi},
 booktitle = {IEEE Information Theory Workshop (ITW), Dublin, Ireland},
 file = {:papers:md_itw10.pdf},
 month = {September},
 note = {},
 pages = {1--5},
 tags = {conf,DetApprox,WiNet,IT,WiNetInfFlow,WiNetErrCorr,Byzantine,SelConf},
 title = {Gaussian diamond network with adversarial jammer},
 type = {4},
 year = {2010}
}

We study the problem where there is a single adversarial node which injects signals to disrupt this communication in a parallel relay network. Like the source, it can only influence the destination through the relays. We develop an approximate characterization of the reliable transmission rate in the presence of such an adversary. A deterministic version of the same problem is solved exactly, yielding insights which are used in the approximate characterization.

@inproceedings{MTDp10,
 abstract = {Lossless transmission of a set of correlated sources
over a deterministic relay network is considered for two traffic
requirements. In distributed multicast, the set of sources are
to be delivered to a set of destinations. The source exchange
requires all the nodes with access to sources to be able to
reconstruct all other sources observed at other nodes. We develop
achievable regions and outer bounds for both these situations. For
linear deterministic networks, these bounds coincide, yielding a
characterization.},
 author = {S. Mohajer and C. Tian and S.N. Diggavi},
 booktitle = {IEEE Information Theory Workshop (ITW) Cairo},
 file = {:papers:mtd_itw10_final.pdf},
 month = {January},
 note = {},
 pages = {},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow,SrcChan,SrcChanSep,NDC,InfExch,SelConf},
 title = {On source transmission over deterministic relay networks},
 type = {4},
 year = {2010}
}

Lossless transmission of a set of correlated sources over a deterministic relay network is considered for two traffic requirements. In distributed multicast, the set of sources are to be delivered to a set of destinations. The source exchange requires all the nodes with access to sources to be able to reconstruct all other sources observed at other nodes. We develop achievable regions and outer bounds for both these situations. For linear deterministic networks, these bounds coincide, yielding a characterization.

@inproceedings{NDc10,
 abstract = {Hierarchical cooperation, introduced by Ozgur, Leveque and Tse showed
that linear scaling is possible in wireless network capacity, asymptotically in network size.
In this paper, we investigate the impact of absence of channel state information
on the performance of hierarchical cooperation. If the product
of coherence time and coherence bandwidth of the wireless
channel is on the order of the number of nodes in the network,
we show that non-coherent hierarchical cooperation achieves
the same order rate as the coherent version. If the product is
smaller than the number of nodes, we show that non-coherent
hierarchical cooperation can still yield sizable gains over multihop
communication, but may not achieve the same order rate
as the coherent version.},
 author = {U. Niesen and S N. Diggavi},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 file = {:papers:ndallerton10final.pdf},
 month = {September},
 note = {},
 pages = {},
 tags = {conf,WiNet,DynamWiNet,NonCohComm,SelConf},
 title = {Non-Coherent Hierarchical Cooperation},
 type = {4},
 year = {2010}
}

Hierarchical cooperation, introduced by Ozgur, Leveque and Tse showed that linear scaling is possible in wireless network capacity, asymptotically in network size. In this paper, we investigate the impact of absence of channel state information on the performance of hierarchical cooperation. If the product of coherence time and coherence bandwidth of the wireless channel is on the order of the number of nodes in the network, we show that non-coherent hierarchical cooperation achieves the same order rate as the coherent version. If the product is smaller than the number of nodes, we show that non-coherent hierarchical cooperation can still yield sizable gains over multihop communication, but may not achieve the same order rate as the coherent version.

@inproceedings{ODp10,
 abstract = {An approximate max-flow min-cut result for arbitrary wireless relay
network was recently established using Gaussian codebooks for
transmission and random mappings at the relays. In this paper,
we show that the approximation result can be established
by using lattices for transmission and quantization along with
structured mappings at the relays. This also extended the original scalar quantizer
analysis to vector quantizers and obtained a slightly better approximation constant.},
 author = {A. Ozgur and S N. Diggavi},
 booktitle = {Proc. of IEEE ISIT 2010, Austin, Texas},
 file = {:papers:od_isit10.pdf},
 month = {June},
 note = {},
 pages = {669--673},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow,SelConf,LatticeChan},
 title = {Approximately achieving Gaussian relay network capacity with lattice codes},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/1005.1284},
 year = {2010}
}

An approximate max-flow min-cut result for arbitrary wireless relay network was recently established using Gaussian codebooks for transmission and random mappings at the relays. In this paper, we show that the approximation result can be established by using lattices for transmission and quantization along with structured mappings at the relays. This also extended the original scalar quantizer analysis to vector quantizers and obtained a slightly better approximation constant.

@inproceedings{PDTc10a,
 author = {E. Perron and S N. Diggavi and I E. Telatar},
 booktitle = {International Zurich Seminar on Communications (IZS)},
 file = {:papers:pdtizs10.pdf},
 label = {pdt_c10a},
 month = {March},
 note = {Invited paper surveying our results on wireless relay network secrecy.},
 pages = {},
 tags = {conf,ITsecrecy,WiNet,IT,WiNetInfFlow,SelConf},
 title = {On information theoretic secrecy for networks},
 type = {4},
 year = {2010}
}

@inproceedings{PDTc10b,
 abstract = {In this paper we consider information-theoretically
secure communication between two special nodes (“source” and
“destination”) in a memoryless network with authenticated relays,
where the secrecy is with respect to a class of eavesdroppers.
We develop achievable secrecy rates when authenticated relays
also help increase secrecy rate by inserting noise into the network.},
 author = {E. Perron and S N. Diggavi and I E. Telatar},
 booktitle = {Proc. of IEEE ISIT 2010, Austin, Texas},
 file = {:papers:pdtisit10.pdf},
 label = {pdt_c10b},
 month = {June},
 note = {},
 pages = {2573 - 2577},
 tags = {conf,ITsecrecy,WiNet,IT,WiNetInfFlow,SelConf},
 title = {On cooperative secrecy for discrete memoryless relay networks},
 type = {4},
 year = {2010}
}

In this paper we consider information-theoretically secure communication between two special nodes (“source” and “destination”) in a memoryless network with authenticated relays, where the secrecy is with respect to a class of eavesdroppers. We develop achievable secrecy rates when authenticated relays also help increase secrecy rate by inserting noise into the network.

@inproceedings{TCDSc10a,
 author = {C. Tian and  J. Chen and S N. Diggavi and S. Shamai},
 booktitle = {Proc. of IEEE ISIT 2010, Austin, Texas},
 month = {June},
 note = {},
 pages = {495--499},
 tags = {conf,SrcChan,ITapprox,NDC,IT},
 title = {Optimality and approximate optimality of source-channel 
separation in networks},
 type = {4},
 year = {2010}
}

@inproceedings{TCDSc10b,
 abstract = {We consider the source-channel separation architecture
for lossy source coding in communication networks. It
is shown that the separation approach is optimal when the
memoryless sources at source nodes are arbitrarily correlated,
each of which is to be reconstructed at possibly multiple
destinations within certain distortions, and the channels in this
network are synchronized, orthogonal and memoryless (i.e., noisy
graphs).},
 author = {C. Tian and  J. Chen and S N. Diggavi and S. Shamai},
 booktitle = {International Conference on Signal Processing and Communications (SPCOM)},
 file = {:papers:tcdsspcom10.pdf},
 month = {June},
 note = {},
 pages = {1--5},
 tags = {conf,SrcChan,ITapprox,NDC,IT,SelConf},
 title = {On source-channel separation in networks},
 type = {4},
 year = {2010}
}

We consider the source-channel separation architecture for lossy source coding in communication networks. It is shown that the separation approach is optimal when the memoryless sources at source nodes are arbitrarily correlated, each of which is to be reconstructed at possibly multiple destinations within certain distortions, and the channels in this network are synchronized, orthogonal and memoryless (i.e., noisy graphs).

@inproceedings{TDc10,
 abstract = {This explores ideas of nearest neighbor search through comparisons. 
The problem is motivated by the problem of navigating an image database with human
assistance. We develop both theoretical bounds, provable algorithms and preliminary
experimental results.},
 author = {D. Tschopp and S N. Diggavi,},
 booktitle = {Information Theory and Applications workshop (ITA), UCSD, San Diego, California},
 file = {:papers:td_ita2010_final.pdf},
 month = {February},
 note = {},
 pages = {1--10},
 tags = {conf,CSalgo,MetricEmb,MClearning,SelConf},
 title = {Facebrowsing: Search and navigation through comparisons},
 type = {4},
 url_arxiv = {http://arxiv.org/abs/0909.2194},
 year = {2010}
}

This explores ideas of nearest neighbor search through comparisons. The problem is motivated by the problem of navigating an image database with human assistance. We develop both theoretical bounds, provable algorithms and preliminary experimental results.

@inproceedings{TDSc10a,
 author = {C. Tian and  S N. Diggavi and S. Shamai},
 booktitle = {Proc. of IEEE ISIT 2010, Austin, Texas},
 month = {June},
 note = {},
 pages = {146--150},
 tags = {conf,SrcChan,NDC,IT},
 title = {The achievable distortion region of bivariate Gaussian source on 
Gaussian broadcast channel},
 type = {4},
 year = {2010}
}

@article{CTDj10,
 abstract = {We characterize the rate region of the two-description multiple description coding for stationary Gaussian sources under the squared error distortion measure.},
 author = {J. Chen and C. Tian and  S N. Diggavi},
 file = {:papers:it_mdstationary_submitted.pdf},
 journal = {IEEE Transactions on Information Theory},
 month = {June},
 note = {},
 number = {6},
 pages = {2868--2881},
 tags = {journal,StatMDC,NDC,IT,MDC},
 title = {Multiple Description Coding for Stationary Sources},
 type = {2},
 volume = {55},
 year = {2009}
}

We characterize the rate region of the two-description multiple description coding for stationary Gaussian sources under the squared error distortion measure.

@article{GFDj09,
 author = {J. Goseling and C. Fragouli and S N. Diggavi},
 journal = {IEEE Communication Letters},
 month = {January},
 note = {},
 number = {1},
 pages = {25--27},
 tags = {journal,InfExch,NetCod},
 title = {Network Coding for Undirected Information Exchange},
 type = {2},
 volume = {13},
 year = {2009}
}

@article{IRADC09,
 author = {K M Z. Islam and P. Rabiei and N. Al-Dhahir and S N. Diggavi and A R. Calderbank},
 file = {:papers:tcommdivembfinal.pdf},
 journal = {IEEE Transactions on Communications},
 month = {June},
 note = {},
 number = {6},
 pages = {1578--1583},
 tags = {journal,DivEmb,OppComm,STcodes},
 title = {Linear Diversity-Embedding STBC : Systems Issues and Applications},
 type = {2},
 volume = {57},
 year = {2009}
}

@article{LCCDj09,
 abstract = {This paper studies how to use the NUM framework in conjunction with diversity
embedded codes to allow multiple streams with different QoS requirements to share
wireless links.},
 author = {Y. Li and M. Chiang and A R. Calderbank and S N. Diggavi},
 file = {:papers:lccdfinal.pdf},
 journal = {IEEE Transactions on Communications},
 month = {May},
 note = {},
 number = {5},
 pages = {1390--1401},
 tags = {journal,DivEmb,OppComm,STcodes,DivEmbNUM},
 title = {Optimal Delay-Rate-Reliability Tradeoff in Networks with Composite Links},
 type = {2},
 volume = {57},
 year = {2009}
}

This paper studies how to use the NUM framework in conjunction with diversity embedded codes to allow multiple streams with different QoS requirements to share wireless links.

@article{TMDj09a,
 abstract = {Approximately solves symmetric Gaussian multiple description problem to within a bit, by sandwiching the MD rate region between two polytopes separated by at most a fraction of a bit.},
 author = {C. Tian and S. Mohajer and  S N. Diggavi},
 file = {:papers:approximatemdregion.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {tmdj09a},
 month = {August},
 note = {},
 number = {8},
 pages = {3869--3891},
 tags = {journal,SymmMDapprox,ApproxIT,NDC,MDapprox,IT},
 title = {Approximating the Gaussian Multiple Description Rate Region Under Symmetric Distortion},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0810.3631},
 volume = {55},
 year = {2009}
}

Approximately solves symmetric Gaussian multiple description problem to within a bit, by sandwiching the MD rate region between two polytopes separated by at most a fraction of a bit.

@inproceedings{JFDj09,
 author = {M. Jafarisiavoshani and  C. Fragouli and S N. Diggavi},
 booktitle = {ACM MobiHoc S3 '09: Proceedings of the 2009 MobiHoc S3 workshop on MobiHoc S3,New Orleans, Louisiana},
 month = {May},
 note = {},
 pages = {21--24},
 tags = {conf,NonCohComm,NetCod,Subsp,CodTh},
 title = {Code construction for multiple sources network coding},
 type = {4},
 year = {2009}
}

@inproceedings{JKAFDp09,
 author = {M. Jafarisiavoshani and L. Keller and A. Argyraki and C. Fragouli and S N. Diggavi},
 booktitle = {IEEE INFOCOM 2009, Rio de Janeiro, Brazil},
 month = {April},
 note = {},
 pages = {2177--2185},
 tags = {conf,SensNet,IDmsg,NetCod},
 title = {Identity aware sensor networks},
 type = {4},
 year = {2009}
}

@inproceedings{JMFDj09,
 author = {M. Jafarisiavoshani and S. Mohajer and C. Fragouli and S N. Diggavi},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Seoul, Korea},
 month = {June},
 note = {},
 pages = {273--277},
 tags = {conf,NonCohComm,NetCod,Subsp,IT},
 title = {On the Capacity of Non-Coherent Network Coding},
 type = {4},
 year = {2009}
}

@inproceedings{KTWj09,
 author = {A. Khisti and S N. Diggavi and G W. Wornell},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Seoul, Korea},
 month = {June},
 note = {},
 pages = {2286--2290},
 tags = {conf,ITsecrecy,KeyGen,IT,WiNetSec},
 title = {Secret key agreement using asymmetry in channel state knowledge},
 type = {4},
 year = {2009}
}

@inproceedings{MDFSj09,
 author = {S. Mohajer and S N. Diggavi and C. Fragouli and D. Tse},
 booktitle = {IEEE Information Theory Workshop (ITW), Volos, Greece},
 month = {June},
 note = {},
 pages = {331--335},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow,RelayInt},
 title = {Capacity of Deterministic Z-Chain Relay-Interference Network},
 type = {4},
 year = {2009}
}

@inproceedings{MDj09,
 author = {S. Mohajer and S N. Diggavi},
 booktitle = {IEEE Information Theory Workshop (ITW), Volos, Greece},
 month = {June},
 note = {},
 pages = {5--9},
 tags = {conf,DetApprox,WiNet,IT,WiNetInfFlow,WiNetErrCorr,Byzantine},
 title = {Deterministic approach to wireless network error correction},
 type = {4},
 year = {2009}
}

@inproceedings{MDTj09,
 author = {S. Mohajer and S N. Diggavi and D. Tse},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Seoul, Korea},
 month = {June},
 note = {},
 pages = {31--35},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow,RelayInt,LatticeChan},
 title = {Approximate Capacity of a Class of Gaussian Relay-Interference Networks},
 type = {4},
 year = {2009}
}

@inproceedings{MJDFj09,
 author = {S. Mohajer and M. Jafarisiavoshani and S N. Diggavi and and C. Fragouli},
 booktitle = {IEEE Information Theory Workshop (ITW), Volos, Greece},
 month = {June},
 note = {},
 pages = {130--134},
 tags = {conf,NonCohComm,NetCod,Subsp,IT},
 title = {On the Capacity of Multisource Non-Coherent Network Coding},
 type = {4},
 year = {2009}
}

@inproceedings{PDTc09d,
 abstract = {This paper studies the idea of noise insertion by authenticated relays through friendly jamming. We develop the secrecy rate achievable in arbitrary (deterministic) networks when there are relays actively helping secrecy.},
 author = {E. Perron and S N. Diggavi and E. Telatar,},
 booktitle = {Information Theory and Applications workshop (ITA), UCSD, San Diego, California},
 file = {:papers:pdtita09final.pdf},
 label = {pdtc09d},
 month = {February},
 note = {},
 pages = {77--84},
 tags = {conf,ITsecrecy,IT,WiNetSec,WiNetInfFlow,ITapprox,WiNet},
 title = {On noise insertion strategies for wireless network secrecy},
 type = {4},
 year = {2009}
}

This paper studies the idea of noise insertion by authenticated relays through friendly jamming. We develop the secrecy rate achievable in arbitrary (deterministic) networks when there are relays actively helping secrecy.

@inproceedings{PDTj09b,
 abstract = {In this paper we find properties that are shared
between two seemingly unrelated lossy source coding setups with
side-information. The first setup is when the source and sideinformation
are jointly Gaussian and the distortion measure is
quadratic. The second setup is when the side-information is an
erased version of the source. We begin with the observation
that in both these cases the Wyner-Ziv and conditional ratedistortion
functions are equal. Next, we consider the case when
there are two decoders with access to different side-information
sources. For the case when the encoder has access to the sideinformation
we establish bounds on the rate-distortion function
and a sufficient condition for tightness. Under this condition,
we find a characterization of the rate-distortion function for
physically degraded side-information. This characterization holds
for both the Gaussian and erasure setups.},
 author = {E. Perron and S N. Diggavi and I E. Telatar},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Seoul, Korea},
 file = {:papers:pdtencsiisit09.pdf|},
 label = {pdt_c09b},
 month = {June},
 note = {},
 pages = {1035--1039},
 tags = {conf,RDSI,NDC,IT,SelConf},
 title = {Lossy Source Coding with Gaussian or Erased Side-Information},
 type = {4},
 year = {2009}
}

In this paper we find properties that are shared between two seemingly unrelated lossy source coding setups with side-information. The first setup is when the source and sideinformation are jointly Gaussian and the distortion measure is quadratic. The second setup is when the side-information is an erased version of the source. We begin with the observation that in both these cases the Wyner-Ziv and conditional ratedistortion functions are equal. Next, we consider the case when there are two decoders with access to different side-information sources. For the case when the encoder has access to the sideinformation we establish bounds on the rate-distortion function and a sufficient condition for tightness. Under this condition, we find a characterization of the rate-distortion function for physically degraded side-information. This characterization holds for both the Gaussian and erasure setups.

@inproceedings{PDTj09c,
 abstract = {We introduce the interference-multiple-access channel,
which is a discrete memoryless channel with two transmitters
and two receivers, similar to the interference channel. One
receiver is required to decode the information encoded at one
transmitter, the other receiver is required to decode the messages
from both transmitters. We provide an inner bound on the
capacity region of this channel, as well as an outer bound for a
special class of such channels. For this class, we also quantify the
gap between inner and outer bound and show that the bounds
match for a semi-deterministic channel, providing a complete
characterization. For the Gaussian case, we show that the gap is
at most 1 bit, yielding an approximate characterization.},
 author = {E. Perron and S N. Diggavi and E. Telatar},
 booktitle = {IEEE International Conference on Communications (ICC), Dresden, Germany},
 file = {:papers:pdt_icc09.pdf},
 month = {June},
 note = {},
 pages = {31--35},
 tags = {conf,ITapprox,WiNet,IT,WiNetInfFlow,IntChan,SelConf},
 title = {On the interference-multiple-access channel},
 type = {4},
 year = {2009}
}

We introduce the interference-multiple-access channel, which is a discrete memoryless channel with two transmitters and two receivers, similar to the interference channel. One receiver is required to decode the information encoded at one transmitter, the other receiver is required to decode the messages from both transmitters. We provide an inner bound on the capacity region of this channel, as well as an outer bound for a special class of such channels. For this class, we also quantify the gap between inner and outer bound and show that the bounds match for a semi-deterministic channel, providing a complete characterization. For the Gaussian case, we show that the gap is at most 1 bit, yielding an approximate characterization.

@inproceedings{PDTp09,
 abstract = {In this paper we consider
secret communication between two special nodes (“source” and
“destination”) in a wireless network with authenticated relays:
the message communicated to the destination is to be kept
information-theoretically (unconditionally) secret from any eavesdropper
within a class. Since the transmissions are broadcast and
interfere with each other, complex signal interactions occur. We
develop cooperative schemes which utilize these interactions in
wireless communication over networks with arbitrary topology,
and give provable unconditional secrecy guarantees.},
 author = {E. Perron and S N. Diggavi and I E. Telatar},
 booktitle = {IEEE INFOCOM 2009, Rio de Janeiro, Brazil},
 file = {:papers:pdtwireless_net_secrecy08.pdf},
 label = {pdt_c09a},
 month = {April},
 note = {},
 pages = {},
 tags = {conf,ITsecrecy,IT,WiNetSec,WiNetInfFlow,WiNet,SelConf},
 title = {On cooperative wireless network secrecy},
 type = {4},
 year = {2009}
}

In this paper we consider secret communication between two special nodes (“source” and “destination”) in a wireless network with authenticated relays: the message communicated to the destination is to be kept information-theoretically (unconditionally) secret from any eavesdropper within a class. Since the transmissions are broadcast and interfere with each other, complex signal interactions occur. We develop cooperative schemes which utilize these interactions in wireless communication over networks with arbitrary topology, and give provable unconditional secrecy guarantees.

@inproceedings{PDTp09a,
 abstract = {The compound wiretap channel generalizes the
classical problem in broadcast information-theoretic secrecy by
allowing a class of potential eavesdroppers. In this paper we
present a new coding scheme that generalizes known approaches
to this problem. The scheme prefixes an artificial multiple
access channel to the transmission scheme in order to design a
structured transmit codebook. The idea is that such a structure
can potentially increase the perfect secrecy rate for the legitimate
users in the presence of the class of eavesdroppers.},
 author = {E. Perron and S N. Diggavi and I E. Telatar},
 booktitle = {IEEE Information Theory Workshop (ITW), Taormina, Italy},
 file = {:papers:pdt09mac_wiretap.pdf},
 month = {October},
 note = {},
 pages = {11--15},
 tags = {conf,ITsecrecy,IT,WiNetSec,SelConf},
 title = {A multiple access approach for the compound wiretap channel},
 type = {4},
 year = {2009}
}

The compound wiretap channel generalizes the classical problem in broadcast information-theoretic secrecy by allowing a class of potential eavesdroppers. In this paper we present a new coding scheme that generalizes known approaches to this problem. The scheme prefixes an artificial multiple access channel to the transmission scheme in order to design a structured transmit codebook. The idea is that such a structure can potentially increase the perfect secrecy rate for the legitimate users in the presence of the class of eavesdroppers.

@inproceedings{SDFPp09,
 abstract = {Abstract—We consider the two message set problem, where a
source broadcasts a common message W1 to an arbitrary set of
receivers U and a private message W2 to a subset of the receivers
P !U . Transmissions occur over linear deterministic channels.
For the case where at most two receivers do not require the
private message, we give an exact characterization of the capacity
region, where achievability is through linear coding.},
 author = {S. Saeedi and S N. Diggavi and C. Fragouli and V. Prabhakaran},
 booktitle = {IEEE Information Theory Workshop (ITW) Taormina, Italy},
 file = {:papers:sdfpitw09.pdf},
 month = {October},
 note = {},
 pages = {406--410},
 tags = {conf,DetApprox,IT,DegMsgSet,OppComm,SelConf},
 title = {On degraded two message set broadcasting},
 type = {4},
 year = {2009}
}

Abstract—We consider the two message set problem, where a source broadcasts a common message W1 to an arbitrary set of receivers U and a private message W2 to a subset of the receivers P !U . Transmissions occur over linear deterministic channels. For the case where at most two receivers do not require the private message, we give an exact characterization of the capacity region, where achievability is through linear coding.

@inproceedings{TDSj09,
 author = {C. Tian and S N. Diggavi and S. Shamai,},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Seoul, Korea},
 month = {June},
 note = {},
 pages = {2477--2481},
 tags = {conf,SrcChan,ApproxIT,NDC,IT},
 title = {An Approximate Characterization For the Gaussian Broadcasting Distortion Region},
 type = {4},
 year = {2009}
}

@article{DCDAj08,
 abstract = {This paper introduces diversity embedded codes, which are space-time codes
that simultaneously give different levels of diversity protection to different information
streams. The paper provides algebraic constructions based on binary codes designed for
rank distance and lifts them to the complex domain. It also gives several example of
linear diversity embedded codes.},
 author = {S N. Diggavi and A R. Calderbank and S. Dusad and  N. Al-Dhahir},
 file = {:papers:dcdafinalversion.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {ddca_j08},
 month = {January},
 note = {},
 number = {1},
 pages = {33--50},
 tags = {journal,DivEmb,OppComm,RankDistCod,CodTh,STcodes},
 title = {Diversity Embedded Space-Time Codes},
 type = {2},
 volume = {54},
 year = {2008}
}

This paper introduces diversity embedded codes, which are space-time codes that simultaneously give different levels of diversity protection to different information streams. The paper provides algebraic constructions based on binary codes designed for rank distance and lifts them to the complex domain. It also gives several example of linear diversity embedded codes.

@article{DDACj08,
 abstract = {This paper develops in a unified way the theory for diversity embedded codes
for broadband (ISI) channels, both from an information theory viewpoint as well as an
algebraic coding perspective. It also develops practical aspects by connecting it to unequal
error protection for image transmission as well as using diversity embedded codes in conjunction with NUM.},
 author = {S. Dusad and S N. Diggavi and N. Al-Dhahir and A R. Calderbank},
 file = {:papers:ddac.pdf},
 journal = {IEEE Journal on Selected Topics in Signal Processing},
 month = {April},
 note = {},
 number = {2},
 pages = {202--219},
 tags = {journal,DivEmb,OppComm,RankDistCod,CodTh,SucRefDiv,DegMsgSet,STcodes},
 title = {Diversity-Embedded Codes : Theory and practice},
 type = {2},
 volume = {2},
 year = {2008}
}

This paper develops in a unified way the theory for diversity embedded codes for broadband (ISI) channels, both from an information theory viewpoint as well as an algebraic coding perspective. It also develops practical aspects by connecting it to unequal error protection for image transmission as well as using diversity embedded codes in conjunction with NUM.

@article{DDCj08,
 abstract = {This paper introduces the problem of diversity embedded codes for inter-symbol
interference (ISI) channels and gives algebraic constructions for them. These constructions
are based on designing binary rank-distance codes with a specific (Toeplitz-like) constraint
and lifting them to the complex domain through set-partitioning maps. As a special case
these algebraic constructions also give rate-diversity optimal codes for transmission over
ISI channels when there are transmit alphabet constraints.},
 author = {S. Dusad and S N. Diggavi and A R. Calderbank},
 file = {:papers:ddcfinalrankdist.pdf},
 journal = {IEEE Transactions on Information Theory},
 month = {November},
 note = {},
 number = {11},
 pages = {4866--4886},
 tags = {journal,DivEmb,OppComm,RankDistCod,CodTh,STcodes},
 title = {Embedded Rank Distance Codes for ISI channels},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0707.4311},
 volume = {54},
 year = {2008}
}

This paper introduces the problem of diversity embedded codes for inter-symbol interference (ISI) channels and gives algebraic constructions for them. These constructions are based on designing binary rank-distance codes with a specific (Toeplitz-like) constraint and lifting them to the complex domain through set-partitioning maps. As a special case these algebraic constructions also give rate-diversity optimal codes for transmission over ISI channels when there are transmit alphabet constraints.

@article{TCDj08,
 abstract = {We study the multi-user successive refinement problem (posed in Pradhan and Ramchandran, 2002), where the users are connected to a central server through links with different noiseless capacities. Each user requires to reconstruct the source in a scalable manner. We provide the best known achievable strategy for the two-user, two-layer case and the complete characterization of the rate-distortion region for the Gaussian source under the mean-squared error (MSE) distortion measure.},
 author = {C. Tian and J. Chen and S N. Diggavi,},
 file = {:papers:multiusersr.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {tcd_j08},
 month = {February},
 note = {},
 number = {2},
 pages = {921--931},
 tags = {journal,RDSI,NDC,RDmultiserSucRef,IT,MDC},
 title = {Multiuser successive refinement and multiple description coding},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0707.4133},
 volume = {54},
 year = {2008}
}

We study the multi-user successive refinement problem (posed in Pradhan and Ramchandran, 2002), where the users are connected to a central server through links with different noiseless capacities. Each user requires to reconstruct the source in a scalable manner. We provide the best known achievable strategy for the two-user, two-layer case and the complete characterization of the rate-distortion region for the Gaussian source under the mean-squared error (MSE) distortion measure.

@article{TDj08,
 abstract = {We introduce the problem of side-information scalable data compression
In this problem, the encoder constructs a progressive description, such that the receiver with high quality side information will be able to truncate the bit-stream and reconstruct in the rate distortion sense, while the receiver with low quality
side information will have to receive further data in order to decode. We provide a complete characterization of the
rate distortion region for the important quadratic Gaussian case with multiple jointly Gaussian side-informations,
where the side information quality does not have to be monotonic along the scalable coding order. Inner and outer bounds are provided to
the rate distortion region for general discrete memoryless sources.},
 author = {C. Tian and S N. Diggavi},
 file = {:papers:siscalable_tiandiggavi.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {td_j08},
 month = {December},
 note = {},
 number = {12},
 pages = {5591--5608},
 tags = {journal,RDSI,NDC,ScalableSIcod,IT},
 title = {Side-information scalable source coding},
 type = {2},
 url_arxiv = {http://arxiv.org/abs/0707.4597},
 volume = {54},
 year = {2008}
}

We introduce the problem of side-information scalable data compression In this problem, the encoder constructs a progressive description, such that the receiver with high quality side information will be able to truncate the bit-stream and reconstruct in the rate distortion sense, while the receiver with low quality side information will have to receive further data in order to decode. We provide a complete characterization of the rate distortion region for the important quadratic Gaussian case with multiple jointly Gaussian side-informations, where the side information quality does not have to be monotonic along the scalable coding order. Inner and outer bounds are provided to the rate distortion region for general discrete memoryless sources.

@article{TSSDj08,
 abstract = {We consider the problem of transmitting a Gaussian
source on a slowly fading Gaussian channel, subject to the mean
squared error distortion measure.  We propose an
efficient algorithm to compute the optimal expected distortion at the
receiver in linear time O(M), when the total number of possible discrete fading
states is M. We also provide a derivation of the optimal
power allocation when the fading state is a continuum, using the
classical variational method.},
 author = {C. Tian and A. Steiner and S. Shamai(Shitz) and S N. Diggavi},
 file = {:papers:distortionbc_tssdfinal.pdf},
 journal = {IEEE Transactions on Information Theory},
 month = {July},
 note = {},
 number = {7},
 pages = {2903--2918},
 tags = {journal,SrcChan,IT},
 title = {Successive Refinement via Broadcasting: Optimizing Expected Distortion of a Gaussian Source over a Gaussian Fading Channel},
 type = {2},
 volume = {54},
 year = {2008}
}

We consider the problem of transmitting a Gaussian source on a slowly fading Gaussian channel, subject to the mean squared error distortion measure. We propose an efficient algorithm to compute the optimal expected distortion at the receiver in linear time O(M), when the total number of possible discrete fading states is M. We also provide a derivation of the optimal power allocation when the fading state is a continuum, using the classical variational method.

@inbook{ZGDMRVWZib08,
 abstract = {This is an initial study into the behavior of traffic matrices at different scales of 
resolution.},
 author = {Y. Zhang and Z. Ge and S N. Diggavi and Z M. Mao and M. Roughan and V. Vaishampayan and W. Willinger and Y. Zhang},
 booktitle = {Festschrift for Thomas Kurtz},
 edition = {Institute of Mathematical Statistics (IMS)},
 file = {:papers:waltertraffic.pdf},
 label = {zgdmrvwz_b08},
 note = {},
 pages = {},
 series = {},
 tags = {BookChap,NetEngr},
 title = {Internet traffic and multiresolution analysis},
 type = {3},
 volume = {},
 year = {2008}
}

This is an initial study into the behavior of traffic matrices at different scales of resolution.

@inproceedings{ADTj08a,
 author = {S. Avestimehr and S N. Diggavi and D N C. Tse, },
 booktitle = {IEEE International Wireless Communications and Mobile Computing Conference (IWCMC), Crete, Greece},
 month = {August},
 note = {},
 pages = {56--61},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow},
 title = {Approximate characterization of capacity in Gaussian relay networks},
 type = {4},
 year = {2008}
}

@inproceedings{ADTj08b,
 author = {S. Avestimehr and S N. Diggavi and D N C. Tse},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Toronto, Canada},
 month = {July},
 note = {},
 pages = {474--478},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow},
 title = {Approximate Capacity of Gaussian Relay Networks},
 type = {4},
 year = {2008}
}

@inproceedings{ADTj08c,
 author = {S. Avestimehr and S N. Diggavi and D N C. Tse},
 booktitle = {IEEE Zurich Seminar on Communications (IZS), Zurich, Switzerland},
 month = {March},
 note = {},
 pages = {92},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow},
 title = {Information flow over compound wireless relay networks},
 type = {4},
 year = {2008}
}

@inproceedings{DDj08,
 author = {S. Dusad and S N. Diggavi,},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Toronto, Canada},
 month = {July},
 note = {},
 pages = {1273--1277},
 tags = {conf,DivEmb,SucRefDiv,STcodes,OppComm,DegMsgSet},
 title = {Successive refinement of diversity for fading ISI MISO channels},
 type = {4},
 year = {2008}
}

@inproceedings{JFDj08,
 author = {M. Jafarisiavoshani and C. Fragouli and S N. Diggavi,},
 booktitle = {IEEE Network Coding Workshop (NetCod), Hong Kong},
 month = {January},
 note = {},
 pages = {1--6},
 tags = {conf,NonCohComm,NetCod,Subsp,NetEngr,NetErrCorr,Byzantine},
 title = {On locating Byzantine attackers},
 type = {4},
 year = {2008}
}

@inproceedings{KDWj08,
 author = {A. Khisti and S N. Diggavi and G W. Wornell,},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Toronto, Canada},
 month = {July},
 note = {},
 pages = {1005--1009},
 tags = {conf,ITsecrecy,KeyGen,IT,WiNetSec,SrcChanSec},
 title = {Secret key generation with correlated sources and noisy channels},
 type = {4},
 year = {2008}
}

@inproceedings{LTDCCj08,
 author = {Y. Li and C. Tian and S N. Diggavi and M. Chiang and A. R. Calderbank},
 booktitle = {IEEE Globecom, New Orleans},
 month = {November},
 note = {},
 pages = {1-6},
 tags = {conf,MDnum,NDC,MDC},
 title = {Network Resource Allocation for Competing Multiple-Description Transmissions},
 type = {4},
 year = {2008}
}

@inproceedings{MDFTj08,
 author = {S. Mohajer and S N. Diggavi and C. Fragouli and D. Tse},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 month = {September},
 note = {},
 pages = {464--474},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow,RelayInt},
 title = {Transmission techniques for relay-interference networks},
 type = {4},
 year = {2008}
}

@inproceedings{MTDj08,
 author = {S. Mohajer and C. Tian and S.N. Diggavi},
 booktitle = {IEEE Data Compression Conference (DCC), Snowbird, Utah},
 month = {March},
 note = {},
 pages = {402--411},
 tags = {conf,AsymMDapprox,ApproxIT,NDC,MDapprox,IT},
 title = {Asymmetric Multilevel Diversity Coding},
 type = {4},
 year = {2008}
}

@inproceedings{PDTj08,
 abstract = {We study a line network with an eavesdropper having access to the
relay transmission. We assume that there could also be a public feedback channel from the destination. We develop achievable strategies to achieve certain rate-equivocation pairs.},
 author = {E. Perron and S N. Diggavi and E. Telatar},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 file = {:papers:pdt_allerton08.pdf},
 month = {September},
 note = {},
 pages = {753--760},
 tags = {conf,ITsecrecy,IT,WiNetSec,WiNetInfFlow,WiNet,InteractiveSec,SelConf},
 title = {Wireless network secrecy with public feedback},
 type = {4},
 year = {2008}
}

We study a line network with an eavesdropper having access to the relay transmission. We assume that there could also be a public feedback channel from the destination. We develop achievable strategies to achieve certain rate-equivocation pairs.

@inproceedings{SDj08,
 author = {A. Sabharwal and S N. Diggavi},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 month = {September},
 note = {},
 pages = {887--894},
 tags = {conf,DetApprox,WiNet,Feedback},
 title = {Compound Gaussian multiple access channels with noisy feedback},
 type = {4},
 year = {2008}
}

@inproceedings{TDGj08,
 author = {D. Tschopp and S N. Diggavi and M. Grossglauser},
 booktitle = {IEEE Zurich Seminar on Communications (IZS), Zurich, Switzerland},
 month = {March},
 note = {},
 pages = {38},
 tags = {conf,WiNet,CSalgo,DynamWiNet,MetricEmb,WiNetRouting},
 title = {Routing in mobile wireless networks},
 type = {4},
 year = {2008}
}

@inproceedings{TMDj08,
 author = {C. Tian and S. Mohajer and S N. Diggavi},
 booktitle = {Information Theory and Applications Workshop, UCSD, San Diego, California},
 month = {January},
 note = {},
 pages = {401--406},
 tags = {conf,SymmMDapprox,ApproxIT,NDC,MDapprox,IT},
 title = {On the Gaussian $K$-description problem under symmetric distortion constraints},
 type = {4},
 year = {2008}
}

@misc{Dp08a,
 author = {Naofal Al-Dhahir and Suhas N. Diggavi and  Anastasios Stamoulis},
 label = {ads_p08a},
 month = {November},
 note = {US Patent  Number US 7,450,656},
 tags = {patent},
 title = {US Patent  Number US 7,450,656: Method for communicating via a channel characterized by parameters that vary in time within a transmission block},
 type = {6},
 year = {2008}
}

@misc{Dp08b,
 author = {Dominique Tschopp and Suhas N. Diggavi and Matthias Grossglauser and Joerg Widmer},
 month = {November},
 note = {European Patent Number EP1883185},
 tags = {patent,DynamWiNet},
 title = {European Patent Number EP1883185: Method and apparatus for beacon management for routing on virtual coordinates},
 type = {6},
 year = {2008}
}

@inbook{CDADCib07,
 author = {J. Chui and S. Das and N. Al-Dhahir and S N. Diggavi and A R. Calderbank},
 booktitle = {Handbook of WiMAX},
 editors = {S. Ahson and M.  Ilyas, CRC Press},
 note = {},
 pages = {},
 series = {},
 tags = {BookChap,STcodes},
 title = {Space-time codes and application in WiMAX},
 type = {3},
 volume = {},
 year = {2007}
}

@inproceedings{ADTj07,
 author = {S. Avestimehr and S N. Diggavi and D N C. Tse},
 booktitle = {IEEE Information Theory Workshop (ITW) Bergen, Norway},
 month = {July},
 note = {},
 pages = {6--11},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow},
 title = {A deterministic model for wireless relay networks and its capacity},
 type = {4},
 year = {2007}
}

@inproceedings{ADTj07a,
 author = {S. Avestimehr and S N. Diggavi and D N C. Tse},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 label = {adt_allerton07a},
 month = {September},
 note = {},
 pages = {},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow},
 title = {Wireless network information flow},
 type = {4},
 year = {2007}
}

@inproceedings{ADTj07b,
 author = {S. Avestimehr and S N. Diggavi and D N C. Tse},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 label = {adt_allerton07b},
 month = {Sept},
 note = {},
 pages = {},
 tags = {conf,DetApprox,ITapprox,WiNet,IT,WiNetInfFlow},
 title = {A Deterministic Approach to Wireless Relay Networks},
 type = {4},
 year = {2007}
}

@inproceedings{CTDj07,
 author = {J. Chen and C. Tian and S N. Diggavi,},
 booktitle = { IEEE Data Compression Conference (DCC), Snowbird, Utah},
 month = {2007},
 note = {},
 pages = {73--82},
 tags = {conf,StatMDC,NDC,IT,MDC},
 title = {Multiple description coding for stationary sources},
 type = {4},
 year = {2007}
}

@inproceedings{DDCj07,
 author = {S. Dusad and S N. Diggavi and A R. Calderbank},
 booktitle = {IEEE Information Theory Workshop (ITW) Bergen, Norway},
 month = {July},
 note = {},
 pages = {32--36},
 tags = {conf,DivEmb,OppComm,RankDistCod,CodTh},
 title = {Rank distance codes for ISI channels},
 type = {4},
 year = {2007}
}

@inproceedings{DMPj07,
 abstract = {We present upper bounds on the capacity of the i.i.d.
binary deletion channel, where each bit is independently deleted
with a fixed probability d. We provide a general approach that
gives a numerical answer for fixed d, and provide an argument
that gives asymptotic upper bounds as d goes to 1. These appear
to be the first non-trivial upper bounds for this probabilistic
deletion channel.},
 author = {S N. Diggavi and M. Mitzenmacher and H D. Pfister},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Nice,France},
 file = {:papers:delubisit07.pdf},
 month = {June},
 note = {},
 pages = {1716--1720},
 tags = {conf,DelChan,IT,SelConf},
 title = {Capacity upper bounds for the deletion channel},
 type = {4},
 year = {2007}
}

We present upper bounds on the capacity of the i.i.d. binary deletion channel, where each bit is independently deleted with a fixed probability d. We provide a general approach that gives a numerical answer for fixed d, and provide an argument that gives asymptotic upper bounds as d goes to 1. These appear to be the first non-trivial upper bounds for this probabilistic deletion channel.

@inproceedings{FMSDj07,
 author = {C. Fragouli and A. Markoupoulou and R. Srinivasan and S N. Diggavi},
 booktitle = {Information Theory and its applications workshop, UCSD, San Diego},
 label = {fmsd_c07},
 month = {January},
 note = {},
 pages = {},
 tags = {conf,NetCod,NetEngr},
 title = {Network monitoring: It depends on your points of view},
 type = {4},
 year = {2007}
}

@inproceedings{JFDGj07,
 author = {M. Jafarisiavoshani and C. Fragouli and S N. Diggavi and C. Gkantsidis},
 booktitle = {ACM SIGCOMM Workshop on Internet Network Management (INM), Kyoto, Japan,},
 label = {jfdg_c07},
 month = {Aug},
 note = {},
 pages = {293--298},
 tags = {conf,NonCohComm,NetCod,Subsp,NetEngr,NetErrCorr},
 title = {Bottleneck discovery and overlay management in network coded peer-to-peer systems},
 type = {4},
 year = {2007}
}

@inproceedings{JFDj07,
 author = {M. Jafarisiavoshani and C. Fragouli and S N. Diggavi},
 booktitle = {IEEE Information Theory Workshop (ITW) Bergen, Norway},
 label = {jfd_c07},
 month = {July},
 note = {},
 pages = {17--21},
 tags = {conf,NonCohComm,NetCod,Subsp,NetEngr},
 title = {Subspace properties of randomized network coding},
 type = {4},
 year = {2007}
}

@inproceedings{LCCDj07,
 author = {Y. Li and M. Chiang and A R. Calderbank and S N. Diggavi},
 booktitle = { IEEE INFOCOM conference, Alaska},
 month = {May},
 note = {},
 pages = {526--534},
 tags = {conf,DivEmb,OppComm,STcodes,DivEmbNUM},
 title = {Optimal Delay-Rate-Reliability Tradeoff in Networks with Composite Links},
 type = {4},
 year = {2007}
}

@inproceedings{PDTj07,
 abstract = {In this paper we study the nested (degraded) message
set problem, where user $i$ requires messages
$W_1, . . . ,W_i$. We study this for a MIMO linear deterministic
broadcast channel model, which is motivated by some recent
successes in using such models to obtain insights into approximate
characterizations for the Gaussian relay and interference
channels. We establish the complete solution for the K = 3
user nested message problem for the MIMO linear deterministic
broadcast channel. We also establish some extremal points for
the general K-user case, where there are only two messages of
interest.},
 author = {V. Prabhakaran and S N. Diggavi and D N C. Tse},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 file = {:papers:degmsgsetbc07.pdf},
 month = {September},
 note = {},
 pages = {},
 tags = {conf,DetApprox,IT,DegMsgSet,OppComm,SelConf},
 title = {MIMO Broadcasting with degraded message sets: A Deterministic Approach},
 type = {4},
 year = {2007}
}

In this paper we study the nested (degraded) message set problem, where user $i$ requires messages $W_1, . . . ,W_i$. We study this for a MIMO linear deterministic broadcast channel model, which is motivated by some recent successes in using such models to obtain insights into approximate characterizations for the Gaussian relay and interference channels. We establish the complete solution for the K = 3 user nested message problem for the MIMO linear deterministic broadcast channel. We also establish some extremal points for the general K-user case, where there are only two messages of interest.

@inproceedings{TDGWj07,
 author = {D. Tschopp and S N. Diggavi and M. Grossglauser and J. Widmer,},
 booktitle = {Information Theory and its applications workshop, UCSD, San Diego},
 file = {:papers:tdgwita2007.pdf},
 month = {January},
 note = {invited paper},
 pages = {124--131},
 tags = {conf,WiNet,CSalgo,DynamWiNet,MetricEmb,WiNetRouting,SelConf},
 title = {Robust Routing for Dynamic Wireless Networks Based onStable Embeddings},
 type = {4},
 year = {2007}
}

@inproceedings{TDj07,
 author = {C. Tian and S N. Diggavi},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Nice,France},
 month = {June},
 note = {},
 pages = {1461--1465},
 tags = {conf,RDSI,NDC,ScalableSIcod,IT},
 title = {On scalable source coding with decoder side informations},
 type = {4},
 year = {2007}
}

@inproceedings{TSSDj07,
 author = {C. Tian and A. Steiner and S. Shamai and S N. Diggavi},
 booktitle = {IEEE Information Theory Workshop (ITW) Bergen, Norway},
 month = {July},
 note = {},
 pages = {42--46},
 tags = {conf,SrcChan,IT},
 title = {Expected distortion for Gaussian source with broadcast transmission strategy over a fading channel},
 type = {4},
 year = {2007}
}

@misc{Dp08,
 author = {Naofal Al-Dhahir and Suhas N. Diggavi and  Anastasios Stamoulis},
 label = {ads_p07a},
 month = {February},
 note = {US Patent Number US 7,173,975},
 tags = {patent},
 title = {US Patent Number US 7,173,975: Estimation of a frequency selective channel with parameters that vary within a transmission block},
 type = {6},
 year = {2007}
}

@misc{RDJPp07,
 author = {Gregory G. Raleigh and Suhas N. Diggavi and  V K. Jones and A. Paulraj},
 label = {rdjp_p07},
 month = {October},
 note = {US Patent Number US 7,286,855},
 tags = {patent},
 title = {US Patent Number US 7,286,855: Method and apparatus for adaptive transmission beamforming in a wireless communication system},
 type = {6},
 year = {2007}
}

@article{DGj06,
 abstract = {This paper was motivated by the transmission of information over finite buffer
channels. Analyzing the capacity of such channels naturally led us to the study of {\em deletion} channels. In deletion channels, in contrast to erasure channels, symbols of the
transmitted sequence are removed, and the receiver does {\em not} know which symbols
have been deleted. This was a classic problem from the 1960s in the context of modeling
synchronization errors. This paper gave the best known (at the time of publication) 
lower bounds to capacity of deletion channels since the 1960s. It also studied erasure channels with arbitrary memory in the erasure process. It showed that even when the erasure
process has memory, feedback does not increase the capacity.},
 author = {S N. Diggavi and M. Grossglauser},
 file = {:papers:ps:finbuf.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {dg_j06},
 month = {March},
 note = {},
 number = {3},
 pages = {1226--1237},
 tags = {journal,DelChan,IT},
 title = {Information transmission over a finite buffer channel},
 type = {2},
 volume = {52},
 year = {2006}
}

This paper was motivated by the transmission of information over finite buffer channels. Analyzing the capacity of such channels naturally led us to the study of \em deletion channels. In deletion channels, in contrast to erasure channels, symbols of the transmitted sequence are removed, and the receiver does \em not know which symbols have been deleted. This was a classic problem from the 1960s in the context of modeling synchronization errors. This paper gave the best known (at the time of publication) lower bounds to capacity of deletion channels since the 1960s. It also studied erasure channels with arbitrary memory in the erasure process. It showed that even when the erasure process has memory, feedback does not increase the capacity.

@inbook{ACDib06,
 abstract = {This was a survey on space-time codes and signal processing with a focus
on recent (when the chapter was written) developments.},
 author = {N. Al-Dhahir and A R. Calderbank and S N. Diggavi},
 booktitle = {MIMO wireless communications},
 edition = {E. Biglieri {\em et al} Cambridge University Press},
 file = {:papers:acdbookchapter.pdf},
 pages = {140--185},
 series = {},
 tags = {BookChap,STcodes},
 title = {Space-time coding for wireless communications: Principles and Applications},
 type = {3},
 volume = {},
 year = {2006}
}

This was a survey on space-time codes and signal processing with a focus on recent (when the chapter was written) developments.

@inbook{Dib06,
 abstract = {This paper brings together how diversity plays a role in diverse topics like
source coding, physical layer wireless communication and mobility in wireless networks.},
 author = {S N. Diggavi},
 booktitle = {Brain and Systems: New Directions in Statistical Signal Processing},
 edition = {Simon Haykin, Jose Principe, Terry Sejnowski, and John McWhirter, MIT Press},
 file = {:papers:divincomm.pdf},
 label = {d_b06},
 pages = {243--286},
 series = {},
 tags = {BookChap,DivInComm,STcodes,MDC,WiNet},
 title = {Diversity in Communication: From source coding to wireless networks},
 type = {3},
 volume = {},
 year = {2006}
}

This paper brings together how diversity plays a role in diverse topics like source coding, physical layer wireless communication and mobility in wireless networks.

@inproceedings{DDCj06,
 author = {S. Dusad and S N. Diggavi and A R. Calderbank},
 booktitle = {IEEE Conference on Information Sciences and Systems, Princeton},
 month = {March},
 note = {invited paper},
 pages = {},
 tags = {conf,DivEmb,OppComm,SrcChan},
 title = {Cross Layer Utility of Diversity Embedded Codes},
 type = {4},
 year = {2006}
}

@inproceedings{DDj06,
 author = {S. Dusad and S N. Diggavi,},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 month = {September},
 note = {},
 pages = {},
 tags = {conf,DivEmb,SucRefDiv,STcodes,OppComm,DegMsgSet},
 title = {On successive refinement of diversity for fading ISI channels},
 type = {4},
 year = {2006}
}

@inproceedings{DTc06,
 abstract = {Diversity embedded codes are opportunistic codes
which take advantage of good channel realizations while ensuring
at least part of the information is received reliably for bad
channels. We establish a connection between these codes and
degraded message set broadcast codes. We characterize the
achievable rate region for the parallel Gaussian degraded message
set broadcast problem, when only the strongest user needs
the private information. Using this, we partially characterize the
set of achievable rate-diversity tuples for the diversity embedded
problem for parallel fading channels. This shows that the diversity-multiplexing
tradeoff for the parallel fading channel is not successively refinable.},
 author = {S N. Diggavi and D N C. Tse},
 booktitle = { IEEE Information Theory Workshop, Uruguay},
 file = {:papers:digtseitw06.pdf},
 month = {March},
 note = {},
 pages = {227--231},
 tags = {conf,DivEmb,SucRefDiv,STcodes,OppComm,DegMsgSet,SelConf},
 title = {On opportunistic codes and broadcast codes with degraded message sets},
 type = {4},
 year = {2006}
}

Diversity embedded codes are opportunistic codes which take advantage of good channel realizations while ensuring at least part of the information is received reliably for bad channels. We establish a connection between these codes and degraded message set broadcast codes. We characterize the achievable rate region for the parallel Gaussian degraded message set broadcast problem, when only the strongest user needs the private information. Using this, we partially characterize the set of achievable rate-diversity tuples for the diversity embedded problem for parallel fading channels. This shows that the diversity-multiplexing tradeoff for the parallel fading channel is not successively refinable.

@inproceedings{FMDj06,
 author = {C. Fragouli and A. Markoupoulou and S N. Diggavi},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 label = {fmd_c06},
 month = {September},
 note = {},
 pages = {},
 tags = {conf,NetCod,NetEngr},
 title = {Topology inference using network coding},
 type = {4},
 year = {2006}
}

@inproceedings{PDTc06,
 author = {E. Perron and S N. Diggavi and I E. Telatar,},
 booktitle = {IEEE International Symposium on Information Theory, Seattle},
 label = {pdt_c06},
 month = {},
 note = {},
 pages = {},
 tags = {conf,RDSI,NDC,IT},
 title = {On the Role of Encoder Side-Information in Source Coding for Multiple Decoders},
 type = {4},
 year = {2006}
}

@inproceedings{TCDj06,
 author = { C. Tian and J. Chen and S N. Diggavi},
 booktitle = {IEEE   Information Theory Workshop (ITW), Chengdu, China},
 month = {October},
 note = {},
 pages = {293--297},
 tags = {conf,RDSI,NDC,RDmultiserSucRef,IT,MDC},
 title = {Multiuser successive refinement and multiple description coding},
 type = {4},
 year = {2006}
}

@inproceedings{TDc06,
 author = {C.Tian and S N. Diggavi},
 booktitle = {Information Theory and Applications Workshop, San Diego},
 month = {February},
 note = {invited paper},
 pages = {},
 tags = {conf,RDSI,NDC,ScalableSIcod,IT},
 title = {On scalable source coding for multiple decoders with side-information},
 type = {4},
 year = {2006}
}

@inproceedings{TDj06,
 author = {C.Tian and S N. Diggavi},
 booktitle = {IEEE International Symposium on Information Theory, Seattle},
 label = {td_j06},
 month = {July},
 note = {},
 pages = {},
 tags = {conf,RDSI,NDC,MultistageWZsucref,IT},
 title = {Multistage successive refinement for Wyner-Ziv source coding with degraded side information},
 type = {4},
 year = {2006}
}

@inproceedings{VDj06,
 abstract = {We investigate source coding in a cascade
communication system consisting of an encoder, a relay
and an end terminal, where both the relay and the
end terminal wish to reconstruct source X with certain
fidelities. Additionally, side-informations Z and Y are
available at the relay and the end terminal, respectively.
The side-information Z at the relay is a physically degraded
version of side-information Y at the end terminal. We give an exact characterization for
the Gaussian quadratic distortion version of the problem and develop inner and outer bounds
for the general case.},
 author = {D. Vasudevan and C.Tian and S N. Diggavi},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 file = {:papers:cascadeallerton06.pdf},
 month = {September},
 note = {},
 pages = {},
 tags = {conf,RDSI,NDC,ScalableSIcod,IT,SelConf},
 title = {Lossy source coding for a cascade communication system with side-informations},
 type = {4},
 year = {2006}
}

We investigate source coding in a cascade communication system consisting of an encoder, a relay and an end terminal, where both the relay and the end terminal wish to reconstruct source X with certain fidelities. Additionally, side-informations Z and Y are available at the relay and the end terminal, respectively. The side-information Z at the relay is a physically degraded version of side-information Y at the end terminal. We give an exact characterization for the Gaussian quadratic distortion version of the problem and develop inner and outer bounds for the general case.

@misc{Dp06,
 author = {Naofal Al-Dhahir and Suhas N. Diggavi and  Anastasios Stamoulis},
 label = {ads_p06},
 month = {October},
 note = {US Patent  Number US 7,130,355 },
 tags = {patent},
 title = {US Patent  Number US 7,130,355: Amelioration of Inter-Carrier Interference in OFDM},
 type = {6},
 year = {2006}
}

@article{BDLMMNVj05,
 abstract = {This studied the online scheduling problem of allocating time or frequency slots
to multiple users in a wireless broadcast channel. The study was from a CS algorithms viewpoint, with criteria such as delay (average/maximal) and {\em stretch} (relative delay).
We gave a  resource-augmented competitive analysis of many scheduling algorithms, and 
demonstrated that a small amount of overprovisioning (resource-augmentation) will make
the scheduling algorithms competitive to the off-line optimal, even for adversarial demand
inputs.},
 author = {L. Becchetti and S N. Diggavi and S. Leonardi and A. Marchetti-Spaccamela and S. Muthukrishnan and T. Nandagopal and A. Vitaletti},
 file = {:papers:dimacs.pdf},
 journal = {Networks},
 label = {ParSchedj05},
 month = {January},
 note = {},
 number = {1},
 pages = {9--22},
 tags = {journal,WiNet,CSalgo,ParSched},
 title = {Downlink Scheduling for Multirate Wireless Networks},
 type = {2},
 volume = {45},
 year = {2005}
}

This studied the online scheduling problem of allocating time or frequency slots to multiple users in a wireless broadcast channel. The study was from a CS algorithms viewpoint, with criteria such as delay (average/maximal) and \em stretch (relative delay). We gave a resource-augmented competitive analysis of many scheduling algorithms, and demonstrated that a small amount of overprovisioning (resource-augmentation) will make the scheduling algorithms competitive to the off-line optimal, even for adversarial demand inputs.

@article{calderbank2005construction,
 author = {Calderbank, Robert and Das, Sushanta and Al-Dhahir, N and Diggavi, Suhas and others},
 journal = {Communications in Information \& Systems},
 label = {cdad_j05},
 number = {1},
 pages = {97--122},
 publisher = {International Press of Boston},
 tags = {journal},
 title = {Construction and analysis of a new quaternionic space-time code for $4 $ transmit antennas},
 type = {2},
 volume = {5},
 year = {2005}
}

@article{DGTj05,
 abstract = {This paper studies the scaling capacity of asymptotically large wireless networks.
It shows that even restricted one-dimensional mobility allows the total capacity of the wireless
network scales linearly with the number of nodes. This implies that per-node capacity scales
as a constant, demonstrating that even restricted mobility can be used to provide non-diminishing throughput for large-scale wireless networks. This is demonstrated by utilizing the multi-user diversity which is instantiated through node-mobility.},
 author = {S N. Diggavi and M. Grossglauser and D N C. Tse},
 file = {:papers:gc.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {dgt_j05},
 month = {November},
 note = {},
 number = {11},
 pages = {3947--3854},
 tags = {journal,WiNet,1Dmob,DynamWiNet},
 title = {One-dimensional mobility increases capacity of wireless adhoc networks},
 type = {2},
 volume = {51},
 year = {2005}
}

This paper studies the scaling capacity of asymptotically large wireless networks. It shows that even restricted one-dimensional mobility allows the total capacity of the wireless network scales linearly with the number of nodes. This implies that per-node capacity scales as a constant, demonstrating that even restricted mobility can be used to provide non-diminishing throughput for large-scale wireless networks. This is demonstrated by utilizing the multi-user diversity which is instantiated through node-mobility.

@article{OSDCj05,
 abstract = {This paper was motivated by non-coherent space-time codes with
fixed transmit alphabet constraints. To obtain maximal diversity order, the subspaces
represented by the space-time codeword matrix needs to be non-intersecting. This led
to the study of how many non-intersecting subspaces can be constructed from a finite
alphabet.},
 author = {F. Oggier and N J A. Sloane and S N. Diggavi and A R. Calderbank,},
 file = {:papers:ps:subsp.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {osdc_j05},
 month = {December},
 note = {},
 number = {12},
 pages = {4320--4325},
 tags = {journal,NonCohComm,STcodes,Subsp},
 title = {Non-intersecting subspaces with finite alphabet},
 type = {2},
 volume = {51},
 year = {2005}
}

This paper was motivated by non-coherent space-time codes with fixed transmit alphabet constraints. To obtain maximal diversity order, the subspaces represented by the space-time codeword matrix needs to be non-intersecting. This led to the study of how many non-intersecting subspaces can be constructed from a finite alphabet.

@inproceedings{CDADj05,
 author = {A R. Calderbank and S.  Das and N. Al-Dhahir and S N. Diggavi},
 booktitle = {Proceedings of IEEE Vehicular Technology Conference, Dallas, Texas, USA},
 month = {September},
 note = {},
 pages = {1791--1795},
 tags = {conf,quaternions,STcodes},
 title = {A Novel Full-Rate Full-Diversity STBC with Application to WiMAX},
 type = {4},
 year = {2005}
}

@inproceedings{CDDMj05,
 abstract = {Inspired by the combinatorial denoising method DUDE, we present efficient algorithms for
implementing this idea for arbitrary contexts or for using it within subsequences. We also propose
effective, efficient denoising error estimators so we can find the best denoising of an input sequence
over different context lengths. Our methods are simple, drawing from string matching methods and
radix sorting. We also present experimental results of our proposed algorithms.},
 author = {S. Chen and S N. Diggavi and S. Dusad and S. Muthukrishnan},
 booktitle = {IEEE   Data Compression Conference (DCC), Snowbird, Utah},
 file = {:papers:strmtch.pdf},
 label = {cddm_c05},
 month = {March},
 note = {},
 pages = {153--162},
 tags = {conf,CSalgo,Denoising,SP,SelConf},
 title = {Efficient String Matching Algorithms for Combinatorial Universal Denoising},
 type = {4},
 year = {2005}
}

Inspired by the combinatorial denoising method DUDE, we present efficient algorithms for implementing this idea for arbitrary contexts or for using it within subsequences. We also propose effective, efficient denoising error estimators so we can find the best denoising of an input sequence over different context lengths. Our methods are simple, drawing from string matching methods and radix sorting. We also present experimental results of our proposed algorithms.

@inproceedings{DADCj05,
 author = {S.  Das and N. Al-Dhahir and S N. Diggavi and A R. Calderbank},
 booktitle = {Proceedings of IEEE Vehicular Technology Conference, Dallas, Texas, USA},
 month = {September},
 note = {},
 pages = {2025--2029},
 tags = {conf,DivEmb,OppComm,STcodes},
 title = {Opportunistic Space-Time Block Codes},
 type = {4},
 year = {2005}
}

@inproceedings{DDCAj05,
 author = {S N. Diggavi and S. Dusad and A. R. Calderbank and N. Al-Dhahir},
 booktitle = {Proceedings of Allerton Conference on Communication, Control, and Computing, Illinois},
 month = {September},
 note = {},
 pages = {},
 tags = {conf,DivEmb,OppComm,RankDistCod,CodTh,SucRefDiv,DegMsgSet},
 title = {On Diversity Embedded codes},
 type = {4},
 year = {2005}
}

@inproceedings{DTj05,
 abstract = {This paper shows that the diversity multiplexing tradeoff for flat fading
channels with a single degree of freedom (MISO/SIMO) is successively refinable.},
 author = {S N. Diggavi and D N C. Tse},
 booktitle = {IEEE International Symposium on Information Theory, Adelaide, Australia},
 file = {:papers:sucrefisit05.pdf},
 month = {September},
 note = {},
 pages = {510--514},
 tags = {conf,DivEmb,SucRefDiv,STcodes,OppComm,DegMsgSet,SelConf},
 title = {Fundamental limits of diversity embedded codes over fading channels},
 type = {4},
 year = {2005}
}

This paper shows that the diversity multiplexing tradeoff for flat fading channels with a single degree of freedom (MISO/SIMO) is successively refinable.

@article{DASCj04,
 abstract = {Spatial diversity is realized through multiple independently fading transmit/receive antenna paths in single-user communications; through independently fading links in multi-user communications; through multiple independent routing paths in networks.
Adopting spatial diversity as a central theme, we study its information-theoretic foundations, then we examine its benefits across the physical (signal transmission/coding and receiver signal processing) and networking (resource allocation, routing, and applications) layers.},
 author = {S N. Diggavi and N. Al-Dhahir and A. Stamoulis and A R. Calderbank,},
 file = {:papers:procieee.pdf},
 journal = {Proceedings of the IEEE},
 label = {dasc_j04},
 month = {February},
 note = {This paper won the  2006 IEEE Donald Fink Prize prize paper award, which is an IEEE-wide paper award across all IEEE journals and magazines.},
 number = {2},
 pages = {217--270},
 tags = {journal,WiNet,ProcIEEE04,FinkAward,STcodes},
 title = {Great Expectations: The value of spatial diversity to wireless networks},
 type = {2},
 volume = {92},
 year = {2004}
}

Spatial diversity is realized through multiple independently fading transmit/receive antenna paths in single-user communications; through independently fading links in multi-user communications; through multiple independent routing paths in networks. Adopting spatial diversity as a central theme, we study its information-theoretic foundations, then we examine its benefits across the physical (signal transmission/coding and receiver signal processing) and networking (resource allocation, routing, and applications) layers.

@inbook{DACib04,
 abstract = {Diversity embedded codes give different levels of diversity protection to 
different information streams. This was the short paper that introduced diversity embedded codes.},
 author = {S N. Diggavi and N. Al-Dhahir and A R. Calderbank},
 booktitle = {AMS edited volume on ``Network Information Theory},
 editors = {P. Gupta G. Kramer and A J. van Wijngaarden},
 file = {{:papers:dimacschap.pdf|}},
 pages = {285--302},
 series = {Series on Discrete Mathematics and Theoretical Computer Science},
 tags = {BookChap,DivEmb,STcodes,OppComm},
 title = {Diversity embedded multiple antenna communications},
 type = {3},
 volume = {66},
 year = {2004}
}

Diversity embedded codes give different levels of diversity protection to different information streams. This was the short paper that introduced diversity embedded codes.

@inproceedings{CDAj04,
 author = {A R. Calderbank and S N. Diggavi and N.Al-Dhahir},
 booktitle = {IEEE International Conference on Communications (ICC), Paris},
 month = {June},
 note = {},
 pages = {483--487},
 tags = {conf,DivEmb,OppComm,RankDistCod,CodTh},
 title = {Space-Time Signaling based on Kerdock and Delsarte-Goethals Codes},
 type = {4},
 year = {2004}
}

@inproceedings{CDDAj04,
 author = {A R. Calderbank and S N. Diggavi and S. Das and N. Al-Dhahir,},
 booktitle = {IEEE International Symposium on Information Theory (ISIT) Chicago},
 month = {June-July},
 note = {},
 pages = {310},
 tags = {conf,quaternions,STcodes},
 title = {Construction and Analysis of a New 4x4 Orthogonal Space-Time Block Code},
 type = {4},
 year = {2004}
}

@inproceedings{DTj04,
 author = {S N. Diggavi and D N C. Tse},
 booktitle = {Allerton Conference on Communication, Control, and Computing, Illinois},
 month = {September},
 note = {},
 pages = {},
 tags = {conf,DivEmb,SucRefDiv,STcodes,OppComm,DegMsgSet},
 title = {On successive refinement of diversity},
 type = {4},
 year = {2004}
}

@inproceedings{DVj04,
 abstract = {We formulate a multi-terminal source coding
problem, where we are required to construct a multiple description
source code for a source sequence when side information
about dependent random processes is available at the decoder
only, or at both the decoder and the encoder. We describe
an achievable rate-distortion region for these problems. In the quadratic Gaussian
case, and when there is common side information among
the decoders, we show that the rate region when both the encoder
and decoder have access to the side information coincides
with that of decoder–only side information. This is analogous
to the single-description (Wyner-Ziv) case, and an explicit
characterization of the rate-distortion region is provided
for this case.},
 author = {S N. Diggavi and V A. Vaishampayan},
 booktitle = {IEEE   Information Theory Workshop (ITW), San Antonio},
 file = {:papers:ps:mdsi.pdf},
 month = {October},
 note = {},
 pages = {},
 tags = {conf,RDSI,NDC,IT,MDC,SelConf},
 title = {On multiple description source coding with decoder side information},
 type = {4},
 year = {2004}
}

We formulate a multi-terminal source coding problem, where we are required to construct a multiple description source code for a source sequence when side information about dependent random processes is available at the decoder only, or at both the decoder and the encoder. We describe an achievable rate-distortion region for these problems. In the quadratic Gaussian case, and when there is common side information among the decoders, we show that the rate region when both the encoder and decoder have access to the side information coincides with that of decoder–only side information. This is analogous to the single-description (Wyner-Ziv) case, and an explicit characterization of the rate-distortion region is provided for this case.

@inproceedings{OSDCj04,
 author = { F. Oggier and N J A. Sloane and S N. Diggavi and A R. Calderbank},
 booktitle = {IEEE International Symposium on Information Theory (ISIT) Chicago},
 month = {June-July},
 note = {},
 pages = {455},
 tags = {conf,NonCohComm,STcodes,Subsp},
 title = {Non-intersecting subspaces with finite alphabet},
 type = {4},
 year = {2004}
}

@article{diggavi2003algebraic,
 author = {Diggavi, Suhas N and Al-Dhahir, Naofal and Calderbank, A. Robert},
 journal = {IEEE Transactions on Information Theory},
 label = {dac_j03},
 number = {10},
 pages = {2403--2414},
 publisher = {IEEE},
 tags = {journal},
 title = {Algebraic properties of space-time block codes in intersymbol interference multiple-access channels},
 type = {2},
 volume = {49},
 year = {2003}
}

@inbook{DACib03,
 author = {S N. Diggavi and N. Al-Dhahir and A R. Calderbank},
 booktitle = {AMS edited volume on Multiantenna Channels: Capacity, Coding and Signal Processing},
 editors = {J. Foschini and S. Verdu},
 series = {Series on Discrete Mathematics and Theoretical Computer Science},
 tags = {BookChap,STcodes,ISIMAC,quaternions},
 title = {Diversity Order of Space-Time Block Codes in Inter-Symbol Interference Multiple-Access Channels},
 type = {3},
 volume = {62},
 year = {2003}
}

@inproceedings{BDj03,
 author = {B. Beferull-Lozano and S N. Diggavi},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Yokohama, Japan},
 month = {July},
 note = {},
 pages = {183},
 tags = {conf},
 title = {Nested trellis codes and shaping for the transmitter side-information problem},
 type = {4},
 year = {2003}
}

@inproceedings{DACj03,
 author = {S N. Diggavi and N. Al-Dhahir and A R. Calderbank},
 booktitle = {IEEE International Conference on Communications (ICC), Anchorage, Alaska},
 month = {May},
 note = {},
 pages = {2643--2647},
 tags = {conf,ISIMAC,STcodes,quaternions},
 title = {Multiuser Joint Equalization and Decoding of Space-Time Codes},
 type = {4},
 year = {2003}
}

@inproceedings{GDMj03,
 abstract = {Implementing end-to-end TCP in wireless networks faces
two problems. First, it is well known that TCP can not distinguish packet losses
due to link failures and that due to network congestion. Second, TCP congestion
control mechanism does not deal effectively with large amount of out-oforder
packet retransmissions; this problem has received less attention in literature.
In this paper, we present solutions to both these problems. In particular,
we present a Link-Layer Header Protection (LHP) protocol which implements
Explicit Loss Notification (ELN) in a simple, scalable manner, addressing the
first problem. We also modify the congestion control mechanism to incorporate
knowledge of ELN and packet loss pattern into retransmission decisions, solving
the second problem. We combine both these solutions with TCP to present
scalable, reliable end-to-end wireless transport protocol.},
 author = {X. Gao and S N. Diggavi and S. Muthukrishnan},
 booktitle = {IEEE International Conference on Communications (ICC), Anchorage, Alaska},
 file = {:papers:lhp.pdf},
 label = {gdm_c03},
 month = {May},
 note = {},
 pages = {66--70},
 tags = {conf,WiNet,TCP,SelConf},
 title = {LHP: An end-to-end reliable transport protocol over wireless data networks},
 type = {4},
 year = {2003}
}

Implementing end-to-end TCP in wireless networks faces two problems. First, it is well known that TCP can not distinguish packet losses due to link failures and that due to network congestion. Second, TCP congestion control mechanism does not deal effectively with large amount of out-oforder packet retransmissions; this problem has received less attention in literature. In this paper, we present solutions to both these problems. In particular, we present a Link-Layer Header Protection (LHP) protocol which implements Explicit Loss Notification (ELN) in a simple, scalable manner, addressing the first problem. We also modify the congestion control mechanism to incorporate knowledge of ELN and packet loss pattern into retransmission decisions, solving the second problem. We combine both these solutions with TCP to present scalable, reliable end-to-end wireless transport protocol.

@inproceedings{VKDKMj03,
 author = {G. Varadhan and S. Krishnan and S N. Diggavi and Y. Kim and D. Manocha},
 booktitle = {Eurographics Symposium on Geometry Processing, Aachen, Germany},
 label = {vkdkm_c03},
 month = {June},
 note = {},
 pages = {116--126},
 tags = {conf,SP,Graphics},
 title = {Efficient Max-norm Distance Computation and Reliable Voxelization},
 type = {4},
 year = {2003}
}

@misc{RDJPp03,
 author = {Gregory G. Raleigh and Suhas N. Diggavi and  V K. Jones and A. Paulraj},
 month = {December},
 note = {US Patent Number US  6,665,545},
 tags = {patent},
 title = {US Patent Number US  6,665,545: Method and apparatus for adaptive transmission beamforming in a wireless communication system},
 type = {6},
 year = {2003}
}

@article{DSVj02,
 abstract = {This designs asymmetric multiple description lattice quantizers, which covers
the entire range of the distortion profile, from symmetric to successive refinement. We present a solution to the labeling problem, which is an important part of the construction. We show that this construction is optimal in the high-rate regime, by comparing it to the information-theoretic bounds.},
 author = {S N. Diggavi and N J A. Sloane and V A. Vaishampayan},
 file = {:papers:ps:final_mdc.pdf},
 journal = {IEEE Transactions on Information Theory},
 month = {January},
 note = {},
 number = {1},
 pages = {174--191},
 tags = {journal,NDC,IT,MDC,LatticeSrc},
 title = {Asymmetric Multiple Description Lattice Vector Quantizers},
 type = {2},
 volume = {48},
 year = {2002}
}

This designs asymmetric multiple description lattice quantizers, which covers the entire range of the distortion profile, from symmetric to successive refinement. We present a solution to the labeling problem, which is an important part of the construction. We show that this construction is optimal in the high-rate regime, by comparing it to the information-theoretic bounds.

@article{FADTj02,
 author = {C. Fragouli and N. Al-Dhahir and S N. Diggavi and W. Turin},
 file = {:papers:ps:mbcjr.pdf},
 journal = {IEEE Transactions on Communications},
 label = {fadt_j02},
 month = {May},
 note = {},
 number = {5},
 pages = {742--753},
 tags = {journal,SP,STWC,STcodes},
 title = {Prefiltered M-BJCR equalizer for MIMO frequency selective channels},
 type = {2},
 volume = {50},
 year = {2002}
}

@inproceedings{ADj02,
 author = {N. Al-Dhahir and S N. Diggavi,},
 booktitle = {Conference on Information Sciences and Systems, Princeton, New Jersey},
 label = {ad_c02},
 month = {March},
 note = {},
 pages = {860--865},
 tags = {conf,WiNet,IT,STWC},
 title = {On achievable rates on time-varying frequency-selective channels},
 type = {4},
 year = {2002}
}

@inproceedings{BDLMMNVj02,
 author = {L.Becchetti and S N. Diggavi and S. Leonardi and A. Marchetti-Spaccamela and S. Muthukrishnan and T. Nandagopal and A.Vitaletti},
 booktitle = {ACM Symp. Par. Alg. and Arch. (SPAA)},
 file = {:papers:ps:spaa02.pdf},
 label = {spaa02},
 month = {August},
 note = {},
 pages = {238--247},
 tags = {conf,WiNet,CSalgo,ParSched},
 title = {Parallel scheduling problems in next generation wireless networks},
 type = {4},
 year = {2002}
}

@inproceedings{DASCj02,
 author = {S N. Diggavi and N. Al-Dhahir and A. Stamoulis and A R. Calderbank},
 booktitle = {Conference on Information Sciences and Systems, Princeton, New Jersey},
 label = {dasc_j02},
 month = {March},
 note = {},
 pages = {259--264},
 tags = {conf,STcodes,DiffSTC,SP,STWC},
 title = {Differential Space-Time Transmission for Frequency-Selective Channels},
 type = {4},
 year = {2002}
}

@inproceedings{DASj02,
 author = {S N. Diggavi and N. Al-Dhahir and A. Stamoulis},
 booktitle = {IEEE International Conference on Communications, New York},
 month = {May},
 note = {},
 pages = {485--489},
 tags = {conf,MIMOofdm,ICIofdm,SP,STWC},
 title = {Inter-carrier interference for MIMO OFDM},
 type = {4},
 year = {2002}
}

@inproceedings{DGj02,
 author = {S N. Diggavi and M. Grossglauser},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Lausanne},
 month = {June},
 note = {},
 pages = {421},
 tags = {conf,DelChan,IT},
 title = {Bounds on capacity of deletion channels},
 type = {4},
 year = {2002}
}

@inproceedings{DGTj02,
 author = {S N. Diggavi and M. Grossglauser and D N C. Tse},
 booktitle = {IEEE International Symposium on Information Theory (ISIT), Lausanne},
 label = {dgt_c02},
 month = {June},
 note = {},
 pages = {352},
 tags = {conf,WiNet,1Dmob,DynamWiNet},
 title = {Even one-dimensional mobility increases capacity of wireless adhoc networks},
 type = {4},
 year = {2002}
}

@inproceedings{SDAj02,
 author = {A. Stamoulis and S N. Diggavi and N. Al-Dhahir},
 booktitle = {IEEE Wireless Communications and Networking Conference, Florida},
 month = {March},
 note = {},
 pages = {465--470},
 tags = {conf,MIMOofdm,ICIofdm,SP,STWC},
 title = {Estimation of fast-fading channels in OFDM},
 type = {4},
 year = {2002}
}

@article{ADj01,
 abstract = {Narrowband interference (NBI) could occur in transmission media such as twisted pair or coaxial cable. We analyzed the effect of such interference on the data throughput for finite-blocklength transmission over noisy inter-symbol interference channels. It was shown that the worst narrowband interference spreads its power over the “sweet spots” of the signal (i.e. where the signal puts highest power). More precisely, the auto-correlation matrix of worst-case narrowband (rank-deficient) interference is shown to have the same eigendirections as the signal. Moreover, if the rank of the covariance matrix of the NBI is M<N (where N is the rank of the signal and M is dimension of the space) then the worst interferer is shown to put its power along the M largest eigendirections of the signal.},
 author = {N. Al-Dhahir and S N. Diggavi},
 file = {:papers:ps:nbi.pdf},
 journal = {IEEE Communication Letters},
 label = {ad_j01},
 month = {June},
 note = {},
 number = {6},
 pages = {233--235},
 tags = {journal,WorstNoise,RobComm,IT},
 title = {Maximum Throughput Loss of Noisy ISI Channels Due to Narrowband Interference},
 type = {2},
 volume = {5},
 year = {2001}
}

Narrowband interference (NBI) could occur in transmission media such as twisted pair or coaxial cable. We analyzed the effect of such interference on the data throughput for finite-blocklength transmission over noisy inter-symbol interference channels. It was shown that the worst narrowband interference spreads its power over the “sweet spots” of the signal (i.e. where the signal puts highest power). More precisely, the auto-correlation matrix of worst-case narrowband (rank-deficient) interference is shown to have the same eigendirections as the signal. Moreover, if the rank of the covariance matrix of the NBI is M

@article{DCj01,
 abstract = {This paper started with a simple question: is the maximum entropy noise the worst noise for additive channels? In the case of scalar channels with a power constraint
on the noise, this is true, as is well known. However, we show that in the vector case, with
covariance constraints, the answer is yes and no. Yes, if the transmit power is large enough
and no otherwise. Along the way we give a solution to the mutual information game with
covariance constraints and show that Gaussian solutions form saddle points, but there could
also be other saddlepoints. We also demonstrate that the information rates can be achieved
using mismatched (Gaussian) decoders.},
 author = {S N. Diggavi and T M. Cover,},
 file = {:papers:ps:worstnoise.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {dc_j01},
 month = {November},
 note = {},
 number = {7},
 pages = {3072--3081},
 tags = {journal,WorstNoise,RobComm,IT},
 title = {Worst additive noise under covariance constraints},
 type = {2},
 volume = {47},
 year = {2001}
}

@article{Dj01,
 abstract = {This is an early study of the capacity of MIMO channels inspired by the work
of Telatar and Foschini. It shows that the capacity grows linearly in the number of transmit
and receive antennas (degrees of freedom) either when the SNR becomes large or when
the number of antennas in the MIMO system becomes large. It shows that the linear rate
growth can be achieved by simpler linear detector structures, such as a matched filter
receiver. However, it shows that the price paid for simple matched filter receivers is that
the SNR growth flattens. It also examines the case where the channel matrix has a 
constant expected Frobenius norm. In this case, even with isotropic fading, it is shown that
the capacity of MIMO channels does not grow linear in the number of antennas, but grows
linearly in SNR. Finally, the effect of time-variation within a transmission block for ISI fading
channels is examined in terms of capacity.},
 author = {S N. Diggavi},
 file = {:papers:ps:fading.pdf},
 journal = {IEEE Transactions on Information Theory},
 label = {d_j01},
 month = {January},
 note = {},
 number = {1},
 pages = {308--325},
 tags = {journal,MIMO,,STcodes,IT,STWC},
 title = {On achievable performance over spatial diversity fading channels},
 type = {2},
 volume = {47},
 year = {2001}
}