TIMING OF CYBER-PHYSICAL ATTACKS ON PROCESS CONTROL SYSTEMS

摘要

This paper introduces a new problem formulation for assessing the vulnerabilities of process control systems. In particular, it considers an adversary who has compromised sensor signals and has to decide on the best time to launch an attack. The task of selecting the best time to attack is formulated as an optimal stopping problem that the adversary has to solve in real time. The theory underlying the best choice problem is used to identify an optimal stopping criterion, and a low-pass filter is subsequently used to identify when the time series of a process variable has reached the state desired by the attacker (i.e., its peak). The complexities associated with the problem are also discussed, along with directions for future research.