GMM-Based Distributed Extended Set-Membership Filtering for Target Tracking Under Cyberattacks

In response to the degradation of target tracking performance in wireless sensor networks (WSNs) under malicious cyberattacks, this article proposes a Gaussian mixture model (GMM)-based distributed robust extended set-membership filtering (ESMF) algorithm to enhance the accuracy of target tracking in the presence of cyberattacks. First, an efficient ESMF-based framework for state estimation of moving targets is presented to estimate the state of each target in an unknown but bounded (UBB) noise environment. Second, a novel GMM-based two-cluster clustering fusion diffusion mechanism is integrated into the ESMF-based framework. By calculating the discrepancy between all local posterior estimates and the fused prior local estimate based on the local interaction of each node, the proposed method effectively screens and fuses local trusted node state estimates. Finally, to alleviate the communication burden and energy consumption in WSN, a distributed filtering architecture has been designed. This architecture enables each node to achieve consensus on target state estimation by exchanging information solely with its neighboring nodes. Experiment results demonstrate the robustness of the proposed method against five common cyberattacks, namely, denial of service (DoS), random, false data injection (FDI), replay, and hybrid attacks.