Dual Rate Control for Security in Cyber-physical Systems

TL;DR

This paper introduces dual rate sampled data controllers that enhance attack detectability in cyber-physical systems by leveraging faster output sampling, even with unbounded disturbances and stealthy attacks.

Contribution

It characterizes how system poles and zeros affect attack detection and proposes multirate controllers that improve security and performance in cyber-physical feedback systems.

Findings

Dual rate controllers detect stealthy attacks effectively.

Faster output sampling improves nominal performance.

Secure sensors and observability are key for attack detection.

Abstract

We consider malicious attacks on actuators and sensors of a feedback system which can be modeled as additive, possibly unbounded, disturbances at the digital (cyber) part of the feedback loop. We precisely characterize the role of the unstable poles and zeros of the system in the ability to detect stealthy attacks in the context of the sampled data implementation of the controller in feedback with the continuous (physical) plant. We show that, if there is a single sensor that is guaranteed to be secure and the plant is observable from that sensor, then there exist a class of multirate sampled data controllers that ensure that all attacks remain detectable. These dual rate controllers are sampling the output faster than the zero order hold rate that operates on the control input and as such, they can even provide better nominal performance than single rate, at the price of higher sampling of the continuous output.