Protocol-Based Distributed Filtering for Stochastic Nonlinear Systems: Tackling Byzantine Attacks and Censored Measurements

This article focuses on the Byzantine-resilient distributed filtering problem for a class of stochastic nonlinear systems (SNSs) over sensor networks. The phenomenon of measurement censoring, which is depicted by a certain Tobit model with both left-censoring and right-censoring thresholds, is considered for the sensor outputs. To save limited network resources, the Round-Robin scheduling mechanism (RRSM) is deployed to govern data transmissions. A Byzantine attack model is introduced with which the Byzantine nodes manipulate the measurement signals arbitrarily before transmitting them to the neighbors. In this research, we endeavor to propose a Byzantine-resilient distributed filtering scheme that ensures an upper bound of filtering error variance and the stochastic boundedness of the resultant filtering error system. Such an upper bound is first calculated in terms of the solutions to coupled difference equations and then minimized by designing suitable filter parameters. Subsequently, the stochastic boundedness analysis issue is addressed for the filtering error dynamics. Finally, the validity of the obtained theoretical results is illustrated by a numerical example.