Ruixin Niu

Jingyang Lu, Ruixin Niu

In this paper, the problem of false information injection attack and defense on state estimation in dynamic multi-sensor systems is investigated from a game theoretic perspective. The relationship between the Kalman filter and the adversary can be regarded as a two-person zero-sum game. Under which condition both sides of the game will reach the Nash equilibrium is investigated in the paper. The multi-sensor Kalman filter system and the adversary are supposed to be rational players. The Kalman filter and the adversary have to choose their respective subsets of sensors to perform system state estimation and false information injection. It is shown how both sides pick their strategies in order to gain more and lose less. The optimal solutions are achieved by solving the minimax problem. Numerical results are also provided in order to illustrate the effectiveness of the derived optimal strategies.

Mengqi Ren, Ruixin Niu

The problem of faulty sensor detection is investigated in large sensor networks where the sensor faults are sparse and time-varying, such as those caused by attacks launched by an adversary. Group testing and the Kalman filter are designed jointly to perform real time system state estimation and time-varying faulty sensor detection with a small number of tests. Numerical results show that the faulty sensors are efficiently detected and removed, and the system state estimation performance is significantly improved via the proposed method. Compared with an approach that tests sensors one by one, the proposed approach reduces the number of tests significantly while maintaining a similar fault detection performance.