Date and Venue: 28-29 Aug 2026, at the SysCon Seminar Room 104, IIT Bombay
Ashish R. Hota ,IIT Kharagpur
Controlling the spread of infectious diseases is a complex challenge that depends heavily on human behavior. When individuals make self-interested choices about adopting costly, partially effective protection, a fundamental tension arises between individual incentives and public health goals. This seminar explores the intersection of epidemiological dynamics, game theory, and information design to model and optimize these behavioral responses within a Susceptible-Infected-Susceptible (SIS) framework. We begin by examining how protection decisions evolve when individuals possess full information, and characterize equilibria of the coupled dynamics of epidemic states and behavioral strategies. We then transition to the more realistic scenario of incomplete information, where individuals must choose whether to protect themselves without knowing their true infection status (e.g., due to asymptomatic transmission). Within this context, we analyze how a principal can use noisy signals and action recommendations to guide rational, utility-maximizing agents towards reducing infection prevalence. Finally, we compare the efficacy of static versus dynamic signaling schemes along an epidemic trajectory, identifying conditions where partial information disclosure paradoxically outperforms full transparency. In summary, this talk provides a comprehensive overview of how strategic signaling can act as a powerful, non-pharmaceutical tool to minimize disease prevalence, while underscoring the delicate balance required when managing strategic populations under uncertainty.
Ashish R. Hota received the B.Tech. and M.Tech. (dual degree) degrees in Electrical Engineering from the Indian Institute of Technology (IIT) Kharagpur, Kharagpur, India, in 2012, and the Ph.D. degree in Electrical and Computer Engineering from Purdue University, West Lafayette, IN, USA, in 2017. He is currently an Associate Professor with the Department of Electrical Engineering, IIT Kharagpur. His research interests include areas of game theory, stochastic optimization, and control of network systems.
Presentation slides will be available after the talk.
Bharadwaj Satchidanandan ,IIT Madras
We consider physical plants controlled by multiple actuators and sensors communicating over a network, where some sensors could be "malicious." A malicious sensor may not report the measurement that it observes truthfully. We address the problem of detecting malicious sensors in a system. We propose a general technique, called "Dynamic Watermarking,'' by which honest actuators in the system can detect the actions of malicious sensors, and disable closed-loop control based on their information
Bharadwaj Satchidanandan is an assistant professor of Electrical Engineering at Indian Institute of Technology Madras. He received his PhD from Texas A&M University where he was advised by Prof. P. R. Kumar, and was a postdoc at Massachusetts Institute of Technology where he was hosted by Prof. Munther Dahleh. His current research interests include Cyber-Physical Systems and their security, power systems, renewable energy integration, mechanism design, and unmanned aerial systems.
Presentation slides will be available after the talk.
Nilabha Saha ,IIT Bombay
We propose an online, data-driven test for detecting adversarial actuator attacks in networked cyber-physical systems modeled as finite-state Markov decision processes (MDPs). The test exploits the ergodicity of the nominal closed-loop MDP: it compares, in total-variation distance, the nominal transition kernel against an empirical occupation measure built from a sliding window of observed state data. The scheme is agnostic to the attack mechanism, assuming only that the adversary's policy is stationary, randomised, and Markovian, and it requires no model of how the attack is carried out. Our main theoretical result is a closed-form detection threshold, governed by the pseudo-spectral gap of the nominal chain, together with a high-probability false-alarm bound obtained from concentration inequalities for Markov chains. We validate the test on a two-queue coupled MDP across a range of adversaries, including denial-of-service and stealthy parameter-drift attacks that fall outside the Markovian assumption, and we characterize the detection boundary as a function of attack intensity and observation-window length. The experiments also expose a structural limit of the approach: attacks that perturb only a small fraction of the state space can stay below the natural fluctuations of the global occupation measure and evade detection.
Nilabha Saha recently completed an integrated dual degree at IIT Bombay, earning a B.Tech. in Computer Science and Engineering and an M.Sc. in Mathematics, along with an Honours in CSE and minors in Applied Statistics and Informatics and in Physics. His research interests lie broadly in cryptography and pseudorandomness, with further interests in arithmetic geometry. He has held research internships in symmetric-key cryptanalysis at EPFL's LASEC and on the cryptanalysis of the extension-field Legendre PRF at KU Leuven's COSIC group. The work presented here, on detecting adversarial attacks in networked Markov decision processes, was carried out in collaboration with Souvik Das, Prof. Debasish Chatterjee, and Prof. Parthanil Roy. He will join Hudson River Trading, Singapore, as an Algo Engineer in September 2026.
Presentation slides will be available after the talk.