Reinforcement Learning-Based Policy Selection of Multi-sensor Cyber Physical Systems Under DoS Attacks

This paper focuses on the problem of optimal policy selection for sensors and attackers in cyber-physical system (CPS) with multiple sensors under denial-of-service (DoS) attacks. DoS attacks have caused tremendous disruption to the normal operation of CPS and it is necessary to assess this damage. The state estimation can reflect the real-time operation status of the CPS and provide effective prediction and assessment in terms of the security of the CPS. For a multi-sensor CPS, different that robust control method is utilized to depict the state of the system against DoS attacks, the optimal policy selection of sensors and attackers is positively analyzed by dynamic programming ideology. To optimize the strategies of both sides, game theory is introduced to study the interaction process between the sensors and the attackers. During the policy iterative optimization process, the sensors and attackers dynamically learn and adjust strategies by incorporating reinforcement learning. To explore more state information, the restriction of state set is loosened, that is the transfer of states are not limited compulsorily. Meanwhile, the complexity of the proposed algorithm is decreased by introducing a penalty in the reward function. Finally, simulation results of the CPS containing three sensors show that the proposed algorithm can effectively optimize the policy selection of sensors and attackers in CPS.