Speaker: Shaunak Mishra
Affiliation: Ph.D. Candidate - UCLA
Abstract: In the past few decades, networked systems have revolutionized, among other things, the way we communicate, and the way we control physical processes. However, addressing the unreliability associated with such networked systems continues to be a fundamental challenge. The unreliability in such networked systems can be attributed to a variety of factors including the lack of coordination among network components, noisy measurements, and security vulnerabilities leading to malicious elements in the network. In view of the factors mentioned above, this dissertation takes steps towards enhancing network reliability by addressing some fundamental challenges in two contemporary network setups: wireless communication networks, and (networked) cyber-physical systems (CPS).
In the context of wireless communication networks, we focus on the unreliability stemming from interference; this is a fundamental bottleneck in reliably scaling data rates. Though recent advances in network information theory have led to a fundamental understanding of interference in wireless networks, the channel models usually assume that interference is always present, and hence do not capture the temporal nature of interference which tends to be bursty in practice. In this context, we develop multicarrier interference management schemes which harness the burstiness of interference. In addition, we develop (tight) impossibility results for a variety of regimes, and hence prove the information theoretic optimality of our interference management schemes in those regimes.
In the context of CPS, we focus on the unreliability arising out of security vulnerabilities. In particular, we study the problem of securing state estimation in linear dynamical systems despite active adversaries. We focus on two attack scenarios: (i) attacks on software/hardware where state estimation is performed (observer attacks), and (ii) attacks on sensors used for state estimation. To protect against observer attacks we propose an architecture where state estimation is performed across multiple computing nodes. To protect against sensor attacks, we propose a Kalman filter based secure state estimation algorithm, and derive (optimal) bounds on the achievable state estimation error given an upper bound on the number of attacked sensors. As a result of independent interest, we give a coding theoretic view of attack detection and state estimation against sensor attacks in a noiseless dynamical system.
Biography: Shaunak Mishra is a Ph.D. candidate in the Electrical Engineering Department at UCLA (advisor: Prof. Suhas Diggavi). He holds a B.Tech degree from the Indian Institute of Technology (IIT) Kharagpur (2010), and an M.S. degree from UCLA (2011). He is a recipient of the Henry Samueli Fellowship (2010-2011), and was a finalist for the Qualcomm Innovation Fellowship 2014. He has held summer internship positions at Yahoo! Labs (scalable machine learning group, 2015), Qualcomm (small cells team, 2013), and EPFL (2012, 2011). Shaunak’s research interests are broadly in information theory and statistics with applications in security, machine learning and wireless networks.
For more information, contact Prof. Suhas Diggavi ()
Date(s) - Jul 25, 2016
10:00 am - 12:00 pm
E-IV Tesla Room #53-125
420 Westwood Plaza - 5th Flr., Los Angeles CA 90095