LASSO-Based Detection and Identification of Actuator Integrity Attacks in Remote Control Systems

This note considers the actuator attack in a remote control system with process and measurement noises. Compared with the sensor attack that has been studied in most existing literature, the malicious attack over actuators can cause more disastrous results due to their direct influence on the system performance. We consider not only detecting the existence of actuator attacks but also identifying the actuators under attack, which is especially important for industrial systems due to the high troubleshooting cost. More specifically, we introduce the notion of prediction residues in Kalman filtering and design a chi(2) attack detector based on the change of residue's probability distribution when the system is under attack. To identify the actuators under attack, we further propose a method based on the least absolute shrinkage and selection operator (LASSO) that characterizes the optimal estimates of sparse actuator attacks in the sense of maximum likelihood and develops an alternating direction method of multipliers (ADMM) scheme for computation. We further design a real-time monitor with low computation and memory complexity and present the pseudo-code for implementation. Numerical simulations and experiments are provided to illustrate the effectiveness of the results.