Secure multi-sensor fusion for systems under deception attacks

This paper studies the design of a fusion estimator for multi-sensor systems subject to deception attacks, where the parameters of deception attacks are not available to the estimator. The difficulties lie in two aspects: (i) how to theoretically determine the proposed fusion estimator is secure; (ii) how to render an insecure fusion estimator secure. To address these problems, at first, attack-parameters estimators are designed, and the estimated parameters are able to converge to the true ones. Based on them, a fusion estimator is designed for an n-dimensional system with m sensors. The main contributions are twofold. (i) A security condition is established. Specifically, it is proved that the fusion estimator is secure, provided that the ratio mn is greater than a certain value. (ii) A sensor-adding algorithm is proposed to make an insecure estimator become secure. Finally, simulations and experiments are presented to demonstrate the effectiveness of the proposed methods.