Jamming attack against remote state estimation over multiple wireless channels: A reinforcement learning based game theoretical approach

In this paper, we design a potential game to investigate the interplay between sensors and attackers in wireless remote estimation systems (RESs), where we consider both closed-loop and open-loop scenarios that depend on whether the sensor's actions are known to the attacker or not. As a typical application of Cyber-Physical Systems (CPSs), wireless RESs have been showing vulnerable to channel jamming attacks. As a countermeasure, the wireless sensor in the estimation system can switch to another independent wireless channel to transmit sensor data. However, a smart attacker can also switch the target channel to launch a jamming attack, resulting in a game between the sensor and attacker. Moreover, the reinforcement learning based methods are proposed to obtain solutions of the designed game models in two scenarios. We also provide extensive simulation results to validate the effectiveness of the proposed methods.