Song Haifang, Ding Derui, Dong Hongli, Han Qing-Long
IEEE Trans Cybern. 2022 May;52(5):3733-3744. doi: 10.1109/TCYB.2020.3016093. Epub 2022 May 19.
This article focuses on the distributed maximum correntropy filtering issue for general stochastic nonlinear systems subject to deception attacks. The considered nonlinear functions consist of a determined one and a stochastic one, and the stochastic signals sent by deception attacks with identified statistic characteristics could be non-Gaussian. The corresponding calculation formulas of both the filter gains and the upper bound of the filter error covariance are proposed by means of the Taylor series expansion and the fixed-point iterative update rule, where the weighted maximum correntropy criterion is utilized to take the place of traditional minimum covariance indexes. Such an upper bound is only dependent on the local information, neighbor information, and the identified statistics of deception attacks and, therefore, the developed filtering scheme realizes the requirement of distributed calculation. Furthermore, a simplified version is obtained by removing weights in the correntropy criterion. Finally, an illustrative example is given to verify the effectiveness of developed distributed maximum correntropy filtering subject to deception attacks.
本文聚焦于遭受欺骗攻击的一般随机非线性系统的分布式最大相关熵滤波问题。所考虑的非线性函数由一个确定性函数和一个随机函数组成，且具有已识别统计特征的欺骗攻击所发送的随机信号可能是非高斯的。借助泰勒级数展开和定点迭代更新规则，提出了滤波器增益和滤波器误差协方差上界的相应计算公式，其中利用加权最大相关熵准则替代传统的最小协方差指标。这样的上界仅依赖于局部信息、邻域信息以及已识别的欺骗攻击统计量，因此，所提出的滤波方案实现了分布式计算的要求。此外，通过去除相关熵准则中的权重得到了一个简化版本。最后，给出一个示例以验证所提出的遭受欺骗攻击的分布式最大相关熵滤波的有效性。
