摘要
There has been a surge of interests in the security of cyber-physical systems(CPSs), yet it is commonly assumed that the adversary has a full knowledge of physical system models. This paper argues that such an unrealistic assumption can be relaxed: the adversary might still be able to identify the system model by passively observing the control input and sensory data. In such a setup, the attack with knowledge of input-output data can be categorized as a Known-Plaintext Attack. A necessary and sufficient condition has been provided, under which the adversary can uniquely obtain the knowledge of the underlying physical system.From the defender's perspective, a secure controller design—which exhibits a low rank structure—is proposed which renders the system unidentifiable to the adversary, while trading off the control system's performance. Finally, a numerical example has been provided to demonstrate the effectiveness of the proposed secure controller design.
基金
supported by the National Natural Science Foundation of China (Grant No. 91748112)
the National Key Research and Development Program of China (Grant No. 2018AAA0101601)。