Towards a Science of Attack Composition, Mitigation and Verification in Cyber Physical Systems-A Passivity Based Approach

Cyber-physical systems (CPS) may be targeted by multiple simultaneous attacks, with the goal of disrupting availability, safety, and performance of the CPS. This project aims to develop a passivity-based framework for modeling, composing, and mitigating attacks on CPS, as well as verifying the models and mitigation. Passivity is an energy dissipation property of dynamical systems that provides basic rules for composition and analysis of interconnected systems. The major activities in this year focused on adaptive, learning based strategies for mitigation in the presence of malicious adversaries. We pursued two different research directions within this broader area. First, we researched and developed passivity-based techniques for learning the unknown parameters of an adversary's attack and adapting the mitigation strategy accordingly. Second, we investigated the CoreMelt denial-of-service attack and proposed passivity-based techniques for analyzing and mitigating the attack.