Dealing with sensor and actuator deception attacks in supervisory control

Abstract

We consider feedback control systems where sensor readings and actuator commands may be compromised by an attacker intending to damage the system. We study this problem at the supervisory layer of the control system, using discrete event systems techniques. The attacker can edit the outputs from the sensors of the system before they reach the supervisory controller as well as it can edit actuator commands before they reach the system. 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 and actuator commands. Intuitively, we search for a supervisor that guarantees the safety of the system even when sensor readings and actuator commands are compromised. Given the similarities of the investigated problem to the standard supervisory control problem, our solution methodology reduces the problem of synthesizing a robust supervisor against deception attacks to a supervisory control problem. This new and intuitive solution methodology improves upon prior work on this topic.