CRII: CPS: Designing Resilient Strategies and Information Structures for Team Games in Cyber-physical Networks

University Of Texas At Dallas, Richardson TX

Investigators

Abstract

This project will develop theoretical and computational tools to assess and improve resilience in cyber-physical networks. Cyber-physical networks are created by the accelerating integration of networked computers and communication devices with physical systems, such as power grids and robotic systems. Although this integration can enhance performance by allowing more sophisticated monitoring and control, it also renders the network vulnerable by increasing the number of access and influence points available to malicious attackers. Red team-blue team scenarios have been used to qualitatively assess and improve security in military and intelligence organizations. The defending (blue) team seeks to operate the network efficiently and securely while the attacking (red) team seeks to disrupt network operation. The research will yield methodology for quantitative red team-blue team scenario analysis based on mathematical models, along with illustrative scenarios arising from problems in power network stability and distributed multi-robot systems. A foundational understanding of resilience and comprehensive mitigation strategies will help to prevent severe attack consequences and protect public health and safety in all application drivers for cyber-physical systems. Despite significant recent research interest, a complete understanding of resilience in cyberphysical networks remains limited. This project will explore a specific game theoretic framework of two-team stochastic dynamic games that mathematically captures crucial issues in red team-blue team scenarios. Specifically, the research will determine how teams of devices in a cyber-physical network can cooperate to make optimal use of information given that (1) each device has only partial information about the state of the network and actions of other devices; (2) the network dynamics and signal flows are subject to random disturbances; and (3) the networks dynamics and signal flows are subject to actions of adversarial teams, whose agents themselves have only partial information about the network. The results will also include algorithms for designing information structures who knows what and when to improve network resiliency using combinatorial optimization and convex relaxation techniques. The theoretical and computational results will be illustrated, validated, and refined by the scenarios in power networks and distributed multi-robot networks, which will be implemented in simulation software and on a multi-robot experimental testbed.