Adaptive CUSUM-Based Residue Analysis for Stealthy Attack Detection in Cyber-Physical Systems

Increased connectivity and sophistication in safety-critical cyber-physical systems (CPS) make them more vulnerable to false data injection (FDI) type attacks. Attackers can leverage the system-agnostic nature of typically used statistical detection methods and smartly launch stealthy FDIs. This work proposes an adaptive CUSUM-based detection method to address this gap. The proposed adaptive CUSUM-based detector can observe and learn from the system behaviour and update its parameters in real-time aiming for fast detection of an FDI raising a minimum number of false alarms.