Sensor attack detection based on active excitation response with uncertain delays

Abstract

Cyber–physical systems face the threat of sensor attacks, which can result in incorrect measurements of the physical state for the controller and lead to erroneous decisions and control instructions. To address this issue, our paper introduces a sensor attack detection method involving the active injection of excitation into the system and a comparison of actual sensor data with the expected response. First, we address challenges posed by sensing noise and uncertain delays by compensating for them and presenting a probability interval derived from the convolution of noise and delay. Second, we create an anticipated response spectrum by adjusting the probability interval within a defined time window under specific active excitation. Third, the disparity between the expected response spectrum and the actual sensor data within the time window is utilized to calculate the probability of a sensor being attacked. Finally, we evaluate the effectiveness of our method through extensive simulations in a vehicle-platoon case study.