Replay Attack Detection for Cyber-Physical Systems with Sensitive States

Abstract

In cyber-physical systems, replay attacks are a type of deception attacks that involve replaying previously recorded sensor data while injecting attack signals into the physical plant. The replay attacks are difficult to detect because the data being replayed is normal. To detect replay attacks, authentication signals are commonly used. However, injecting authentication signals may affect the performance of system states that are sensitive to noise or disturbance. To address this issue, the effect of authentication signals on sensitive states is measured in the sense of mean square error, and conditions that the authentication signal has a limited effect or even no effect on sensitive states are presented. Combining these conditions, optimization problems are constructed to design the authentication signal. Finally, the proposed method is validated through simulation results.