Data-Driven Stealthy Sparse Attack Detection and Identification for Cyber Physical Systems

This paper investigates the issues of attack detection and identification in Cyber-Physical Systems (CPSs) under sparse sensor and actuator attacks, focusing on systems with unknown parameters. Initially, stealthy sparse attacks are formulated utilizing input-output information and subspace identification techniques, and the most destructive attacks are established based on an attack impact index. Subsequently, a data-driven attack detection strategy is proposed based on coding principles, and a Gated Transformer is employed to identify the compromised sensors and actuators. Finally, the proposed methodologies are validated through simulations on a supersonic aircraft, with the coding-based attack detector integrated on Zynq-7000 FPGA.