Optimal denial-of-service attack on feedback channel against acknowledgment-based sensor power schedule for remote estimation

This paper considers a class of malicious attacks against remote state estimation in cyber-physical systems. An energy-limited sensor adopts an acknowledgment-based online power schedule for remote state estimation. A malicious attacker is assumed to be able to eavesdrop the acknowledge signal and launch a Denial-of-Service attack to jam the feedback channel, thereby misleading the sensor with consecutive energy consumption. With the capability constraint, the attacker aims at seeking an optimal policy which maximizes a long-Term average reward function consisting of both estimation error and energy consumption. We formulate the problem as a Markov decision process and prove the existence of the optimal stationary and deterministic policy. The optimal policy is then shown to have a threshold structure, which significantly reduce the computational complexity. Simulation examples are provided to demonstrate the analytical results.