GMM-Based Distributed Kalman Filtering for Target Tracking Under Cyberattacks
Jingxian Luo, Hongbo Zhu
Abstract
To address the target tracking problem in wireless sensor networks subject to malicious cyberattacks, this letter proposes a distributed Kalman filtering approach based on the Gaussian mixture model (GMM). In the estimation process, a GMM-based information fusion scheme is introduced between the measurement correction step and the time-update step. For each node, the scheme uses GMM to cluster the node and its adjacent nodes into two sets and classifies the two sets into trust set and untrust set according to majority voting. The posterior estimate is refined by fusing local estimates of all nodes in the trust set. Simulation results show that the proposed approach tracks targets effectively under random attack and false data injection attack. Furthermore, when compared with the k-means-based distributed Kalman filtering approach, the GMM-based one is more robust. Finally, the performance of the approach under hybrid attack is discussed in order to consider the stability of the system. We still get better test results under more stringent conditions.