This paper is concerned with the multitarget tracking over decentralized sensor networks where the network can potentially be compromised by malicious cyber attacks. We consider the hybrid cyber attacks, including denial of service (DoS), false data injection (FDI), and extra packet injection (EPI) attack. We first establish the feature model of DoS, FDI and EPI attacks for decentralized multitarget tracking. Then, we propose a decentralized multitarget tracking algorithm against DoS, FDI and EPI attacks, which consists of three phases: prediction, adaptation and combination. The adaptation phase is to update the estimate of each node with its own measurements and all its neighbors' measurements. The combination phase is to fuse the estimate of each node with all its neighbors' estimates. By incorporation of the neighbors' measurements and fusing the neighbors' estimates, it can dramatically reduce the adverse effect of cyber attacks and provide reliable tracking performance. Numerical experiments are provided to demonstrate the effectiveness of the proposed algorithm.

本文关注分散式传感器网络上的多目标跟踪，其中网络可能会受到恶意网络攻击的危害。我们考虑混合网络攻击，包括拒绝服务 (DoS)、虚假数据注入 (FDI) 和额外数据包注入 (EPI) 攻击。我们首先建立了 DoS、FDI 和 EPI 攻击的特征模型，用于去中心化多目标跟踪。然后，我们提出了一种针对 DoS、FDI 和 EPI 攻击的分散式多目标跟踪算法，该算法包括三个阶段：预测、适应和组合。适应阶段是用它自己的测量值和它所有邻居的测量值更新每个节点的估计。组合阶段是将每个节点的估计与其所有邻居的估计融合。通过合并邻居的 测量并融合邻居的估计，它可以显着减少网络攻击的不利影响并提供可靠的跟踪性能。提供了数值实验来证明所提出算法的有效性。