Data-driven attack policy design for cyber-physical systems under channel constraints

This paper is concerned with the design of data -driven sensor injection attack for cyber-physical systems (CPSs) under channel constraints, where the attacker can only access a part of the sensor channels and the accessible channels switch over time. To enhance the attack's effectiveness while maintaining its stealthiness, the design problem is formulated as a constraint -type L 2 - gain optimization problem. Then, by optimizing the attack -direction matrix using the data -driven parametrization method, an attack policy with channel constraints is proposed. Specifically, the necessary and sufficient design conditions are established in terms of the attack stealthiness. Furthermore, the optimized attack stealthiness index and attack effectiveness index under channel constrains are obtained, and it is theoretically proven that the attack performance is reduced due to the passive channel switching. The effectiveness of the data -driven attack policy is illustrated by the IEEE 6 bus power system.