Robust Secure Distributed Estimation Based on Statistical–Temporal Fusion Over Adversarial Sensor Networks Against Noisy Interference

This article studies the robust secure distributed estimation over adversarial wireless sensor networks, where input measurement noises and output impulsive noises are considered. In this case, the existing distributed algorithms suffer from severe performance degradation. Hence, we derive a new adversarial network model to address these issues and propose a robust statistical-temporal combination based secure diffusion bias compensated maximum correntropy criterion (RSTCS-dBCMCC) algorithm. Firstly, to eliminate the influence of bias induced by input measurement noises, the dBCMCC algorithm is introduced under the adversarial network environment with impulsive noisy interference. To further mitigate adversarial attacks, a robust statistical-temporal fusion based multi-dimensional attack detection method is designed, in which spatial temporal combination based reference selection is proposed, and kernel adaptive adjustment with statistical information combination based kernel risk-sensitive loss (SC-KRSL) function is presented to update the temporal combination factor. The theoretical performance of the proposed algorithm is also analyzed in detail. Finally, simulations demonstrate that our proposed algorithm exhibits both superior performance and robustness when compared to state-of-the-art secure distributed algorithms under complex attack scenarios and impulsive interference environments.