This project seeks to develop design methodologies for the synthesis of cyber-physical systems (CPS) that verifiably satisfy given safety and performance requirements when an unknown set of system components is compromised. The need for such design methodologies is exemplified by recent intrusions into nuclear facilities and ransomware attacks on municipal governments, in which adversaries found weak points in cyber defenses that were leveraged to control safety-critical physical infrastructures. The research plan is grounded on two application scenarios: (i) a group of unmanned vehicles that must complete high-level task objectives while avoiding collisions in the presence of false and malicious sensor and control inputs, and (ii) a smart building in which IoT apps send malicious commands to the building HVAC and other safety-critical systems.
The PI will develop algorithms to compute control policies in the presence of attacks that inject arbitrary sensor measurements or control signals, disrupt availability of sensor or control messages, and/or modify controller set points. The first research thrust will investigate and develop control strategies for safety and reachability of nonlinear systems under attack by extending the notions of control barrier and control Lyapunov functions to adversarial settings. The second thrust will investigate resilient synthesis of more complex task specifications using the control algorithms of thrust one as building blocks. The PI will develop novel approaches to model adversarial cyber-physical interactions as stochastic games by developing resilient finite-state abstractions of nonlinear systems. Finite-state control policies will be developed by approximating the game solutions. This thrust will investigate contract-based decomposition algorithms for solving the games in a distributed system with multiple (potentially malicious) decision-making agents. Each thrust of the project will be validated through experimentation and testing on two custom platforms, namely, a multi-robot testbed and a smart building simulation framework. This project will result in models and algorithms to improve safety, performance, and security of CPS including connected and autonomous vehicles, industrial control systems, intelligent traffic management systems, medical devices, and manufacturing CPS. The PI will develop “serious games†to enhance public interest while providing insight into human decision-making. Algorithms for secure control developed in the project will be experimented on by undergraduate capstone students under the supervision of the PI’s graduate students.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
