Seminar in Signals and Systems on Networked Systems

Speaker: Mohit Daga

Title: Stability of Continuous Gossip in Networked Systems under Byzantine and Non-Byzantine Behavior

Abstract: Gossip-based information spreading is a fundamental primitive in networked systems, with connections to distributed control, consensus, and information flow over graphs. In this talk, I consider a continuous gossip model in which information arrives over time and is propagated through local pairwise interactions. The main focus is on the stability of such systems: under what conditions can the network continuously absorb and disseminate incoming information without the backlog growing unbounded?

I will present results for both complete graphs and general graph topologies, and compare the behavior of systems with only honest agents to those with Byzantine actors that may deviate arbitrarily from the protocol. The perspective is that of a dynamical system evolving over a network, where graph structure, communication constraints, and adversarial behavior jointly determine long-term performance. In particular, I will discuss how topology affects stability margins, how Byzantine nodes alter the achievable operating region, and what these results suggest for robust information flow in adversarial networked systems.

Speaker Biography: Mohit Daga is a final-year PhD student in Computer Science at KTH Royal Institute of Technology. His research focuses on algorithm design under locality constraints, with current work on the stability of continuous gossip processes in networks, particularly under adversarial conditions. In addition to his PhD Studies, he works as a Master’s thesis supervisor at KTH.

During his PhD, he has been a visiting researcher at the University of Hawai‘i at Mānoa (USA), the National Institute of Informatics (Japan), CyStar Labs at IIT Madras (India), and the IMDEA Software Institute (Spain). Mohit holds a Master’s degree in Computer Science from IIT Madras, India.