Security Analysis and Control Under Periodic DoS Attacks.

This paper is focused on security analysis and control problems of cyber-physical systems under denial-of-service (DoS) attacks, which jam the controller-actuator channel. Considering the limited attack energy of malicious adversaries, DoS attacks are described by a periodic model. Different from existing results, both the security analysis and secure controller design problems are addressed based on the characteristics of DoS attack model rather than utilizing the switching system theory to solve the aforementioned problems. Firstly, sufficient conditions are derived to ensure that the resultant closed-loop cyber-physical system under DoS attacks can preserve the exponential stability, and the critical value of the attack period is derived, below which the stability is deteriorated. Secondly, the relation between our proposed conditions and reinforcement learning based control is established, based on which the security of reinforcement learning based control can be evaluated effectively. Lastly, both DoS attacks and external disturbances are considered in a unified framework, and sufficient conditions are proposed to evaluate the security of the closed-loop cyber-physical systems and design a secure controller. Finally, a mobile robot is adopted to validate the efficacy of the proposed methods.