Passivity Framework for Modeling, Composing and Mitigating Cyber Attacks

Thesis (Ph.D.)--University of Washington, 2016-08

Cyber systems form the backbone of our society, serving infrastructures for health, energy, transportation, and finance to name a few. As the reliance on cyber systems grows, the impact of cyber attacks also increases. Recent cyber incidents demonstrate that cyber attacks result in not only financial cost, but also compromise the safety of critical infrastructures. Emerging cyber threats including advanced persistent threats (APT) show growing sophistication of attackers. Attackers exploit large number of entry points with different vulnerabilities and adaptively change attack strategies based on observed information of the targeted system. These features are captured by existing adversary models. To defend against such emerging threats, a new approach is needed for modeling and mitigating cyber attacks. The goal of this thesis is providing fundamental approaches toward addressing these challenges. In this thesis, we study control and game theoretic approaches, both of which are developed under the passivity framework. Using dynamical systems theory, we model adaptive and time-varying dynamics of cyber attacks. We develop passivity-based composition rules that characterizes the impact of multiple simultaneous attacks, and design mitigation strategies against adversary models using passive structures. We have developed this approach for attacks including wormhole, jamming and malware propagation. Patrolling and resource takeover games are also studied under passivity framework.