Detection of covert attacks and zero dynamics attacks in cyber-physical systems

A new approach to detect covert attacks and zero dynamics attacks on cyber-physical systems is presented in this paper. These attack strategies are very sophisticated and rely on perfect system knowledge. The resulting attack signal is completely invisible in the sensor readings. Hence an ordinary fault diagnosis system is not able to detect such attacks and to trigger an alarm. Our goal is to insert a modulation matrix in the path of the control variables. By altering the input behavior of the process, the adversary will lose his perfect system knowledge and therefore his attacks can be revealed. We design the modulation matrix by focusing on covert attacks. By guaranteeing that the modulation changes the input directions, we unmask zero dynamics attacks as well, since multivariable zeros depend on frequency and direction.