Identification of Periodic Sensor-Reading Modification Attacks in Cyber-Physical Systems

Cyber-physical systems are extensively utilized in industry and critical infrastructures due to their ability to facilitate mutual cooperation and interaction between physical and virtual worlds. Nevertheless, their security has been increasingly threatened by cyber-attacks. Motivated by the need to address the resulting security concerns, this work studies a class of periodic sensor attacks that periodically modify sensor-readings to disrupt the performance of remote estimators in a system. After analyzing system behavior under such attacks based on discrete event systems, we focus on identifying malicious modifications on each event. In particular, we formalize the concept of modification identifiability under attacks, which indicates whether any modification can be uniquely identified in finite observations. Its verification problem is addressed by developing a novel state estimator.