Detection of Data Integrity Attacks in Distributed State Estimation

We study the security issue of distributed state estimation under data integrity attacks over wireless sensor networks. We design a detector based on statistical learning to judge the compromised estimate sent from the neighboring sensors. To obtain the best estimation performances, we find an optimal estimator for sensors equipped with the malicious data detector, and find a sufficient condition to ensure the stability of the trace of estimation error covariances (EECs). In addition, we explore the relationship between the steady-state EEC and the parameters of the detector. Finally, by numerical simulations, we show the performances of several typical detectors proposed in the existing works, and verify the influence of the detector parameters on the estimation performances.

Automated summary

This study investigates the security issue of distributed state estimation under data integrity attacks in wireless sensor networks. A detector based on statistical learning is designed to detect compromised estimates from neighboring sensors. Optimal estimator and stability condition for estimation error covariances are found for sensors equipped with the malicious data detector. The relationship between steady-state EEC and detector parameters is explored, and the performances of various detectors are verified through numerical simulations.