A formal verification approach to revealing stealth attacks on networked control systems

Abstract

We develop methods to determine if networked control systems can be compromised by stealth attacks, and derive design strategies to secure these systems. A stealth attack is a form of a cyber-physical attack where the adversary compromises the information between the plant and the controller, with the intention to drive the system into a bad state and at the same time stay undetected. We define the discovery problem as a formal verification problem, where generated counterexamples (if any) correspond to actual attack vectors. The analysis is entirely performed in Simulink, using Simulink Design Verifier as the verification engine. A small case study is presented to illustrate the results, and a branch-and-bound algorithm is proposed to perform optimal system securing.