Moving-Target Defense against Controller Attacks in Cyber-Physical Systems

In the proposed project, we will develop novel moving target defenses against compromised controllers in time- and safety-critical Cyber-Physical Systems (CPSs).

First, we develop a dynamical-systems-based modeling framework to characterize all possible stealthy controller attacks, assuming the anomaly detection system is trustworthy.

We then use the characterization to model the interaction between compromised controllers and detectors as an attacker-defender game, where the detectors´ actions consist of reconfiguring various system parameters.

By computing mixed strategy Nash equilibria, we obtain optimal randomized defense policies to mitigate the effects of stealthy controller attacks.

We will recruit a doctoral student to carry out the research program, under the supervision of a postdoc and the project leader (both funded in part by the project). To solve the posed security problem, tools from control, optimization, and game theory will mainly be used.

Time-critical CPSs, such as power systems and industrial control systems, have in the past decade become the targets of cyber-attacks with possibly devastating physical effects.

We argue the considered controller attack scenario has been somewhat overlooked in the related research literature, but it is a particularly important and possible scenario.

Traditional security solutions are often not feasible for the systems considered, and thus our proposed application-layer solution is an attractive and novel way forward.