Secure Control for Cyber-Physical Systems Under Malicious Attacks

This article investigates the secure control problem for cyber-physical systems when malicious data are injected into cyber realm, which directly connects to actuators. Based on moving target defense (MTD) and reinforcement learning, we propose a novel proactive reactive scheme. First, system $(A,B)$ is modeled as switching consisting of several controllable pairs notation="LaTeX">$(A,\mathcal {B}_{l})$ facilitate construction MTD The can be altered update dynamics under certain unpredictable probabilities each subsystem, prevent adversaries from effective attacks. Second, both attack detection isolation schemes designed accurately locate exclude compromised actuators sequence. Third, learning algorithm based zero-sum game theory proposed design scheme there exist no subsystems switch. To demonstrate effectiveness scheme, three-tank unknown attacks illustrated.