Towards resilient supervisors against sensor deception attacks

Abstract

We consider feedback control systems where sensor readings may be compromised by a malicious attacker intent on causing damage to the system. We study this problem at the supervisory layer of the control system, using discrete event systems techniques. We assume that the attacker can edit the outputs from the sensors of the system before they reach the supervisory controller. In this context, we formulate the problem of synthesizing a supervisor that is robust against a large class of edit attacks on the sensor readings. We solve this problem using a new methodology that improves upon prior work on this topic. The solution methodology is based on the solution of a partially observed supervisory control problem with arbitrary control patterns. We also provide results on the existence of a supremal robust supervisor.

Synthesis of Supervisors Robust Against Sensor Deception Attacks

“…However, robustness in the previous literature is related to communication delays [8], [9], loss of information [10], or model uncertainty [4], [5], [7], [6]. Exceptions to that are [11], [12], [13], [14], where the problem of synthesizing supervisors robust against attacks was investigated. The results of [13] are related to actuator deception attacks.…”

“…In [11], [12], [13], [15], [14], the problem of synthesizing supervisors robust against attacks was investigated. Our work differs from [11], [12], [14] as we provide a general gametheoretical framework that solves the problem of synthesizing supervisors robust against general classes of sensor deception attacks.…”

“…In [11], [12], [13], [15], [14], the problem of synthesizing supervisors robust against attacks was investigated. Our work differs from [11], [12], [14] as we provide a general gametheoretical framework that solves the problem of synthesizing supervisors robust against general classes of sensor deception attacks. The solution methodology in [11], [12], [14] follows the standard supervisory control solution methodology, where only results about one robust supervisor against a specific class of sensor deception attacks is provided.…”

“…Our work differs from [11], [12], [14] as we provide a general gametheoretical framework that solves the problem of synthesizing supervisors robust against general classes of sensor deception attacks. The solution methodology in [11], [12], [14] follows the standard supervisory control solution methodology, where only results about one robust supervisor against a specific class of sensor deception attacks is provided. Conditions on the existence of robust supervisors against a possible set of sensor deception attacks with a normality condition on the plant are provided in [11].…”

“…However, the supervisory control framework in [15] differs from the standard framework since the authors assume that the supervisor can actively change the state of the physical process. Finally, [14] provides a methodology to synthesize a maximal controllable and observable supervisor against unbounded sensor deception attacks.…”