Detection and Identification of CPS Attacks with Application in Vehicle Platooning: a Generalized Luenberger Approach

This paper considers the security of distributed Cyber-Physical Systems (CPS) under malicious attacks, where the CPS components interact over a communication network. We consider the deception attacks both at the individual component level as well as the network level. At the individual level, false-data-injection corrupts sensor measurements and/or actuator signals. While at the network level, the communication between CPS components may be altered by adversarial false-data. We develop a robust model-based detection and identification algorithm for a class of discrete Linear Time Invariant (LTI) systems with delays via a Luenberger-like observer termed Generalized Luenberger Observer (GLO). The optimal GLO provides a tight bound for the residue between the monitor signals and their estimation so that it could distinguish the attack signals from intrinsic bounded noises and modeling uncertainties. Furthermore, a structurally constrained optimal GLO could also identify the place where such CPS attacks take place. Finally, we apply it to an application of the longitudinal platoon with four vehicles.