Data-Driven Model Generation for Deception Defence of Cyber-Physical Environments

Cyber deception is a burgeoning defence technique that provides increased detection and slowed attack impact. Deception could be a valuable solution for defending the slow-to-patch and minimally cryptographic industrial Cyber-Physical Systems. However, it is necessary for cyber-physical decoys to appear connected to the physical process of the defended system to be convincing. In this paper, the authors present a machine-learning approach to learn good-enough models of the defended system to drive realistic decoy response. The results of studying this approach with simulated and real building systems are discussed.