Privacy-Preserving Distributed Kalman Filtering Based on State Decomposition and Dynamic Mask
Kang Si, Peng Li, Zhi-Peng Yuan, Xiangdong Jiang, Zheng-Xian Wei
Abstract
In recent years, distributed Kalman filtering has emerged as a critical method for state estimation over wireless sensor networks. However, the sharing of information among various nodes poses privacy challenges, potentially leading to data leakage. The required scheme can not only effectively protect the privacy of the information but also ensure the estimation performance. Toward this goal, a privacy-preserving distributed filtering scheme is designed by blending the state decomposition approach with the dynamic mask mechanism. Specifically, the original local state estimation is first randomly decomposed into the private and public states, where the private state is updated locally, remaining invisible to other nodes, while only the public state is shared with neighbors. In order to further improve the privacy, the dynamic masking technique is adopted, in which the dynamic affine mask is inserted to the exchanged public state. In addition, we provide a detailed analysis of the convergence, estimation performance, and privacy properties of the proposed algorithm. Finally, simulations are performed to demonstrate the validity of our developed privacy-preserving distributed algorithm.