基于异常值的拟态裁决优化方法  

Mimic ruling optimization method based on executive outliers

作　　者：高振斌 贾广瑞 张文建 谭力波 Gao Zhenbin;Jia Guangrui;Zhang Wenjian;Tan Libo(School of Electronic&Information Engineering,Hebei University of Technology,Tianjin 300401,China;Information Technology Research Institute,PLA Strategic Support Force Information Engineering University,Zhengzhou 450002,China;Information Technology Innovation Center of Tianjin Binhai New Area,Tianjin 300450,China)

机构地区：[1]河北工业大学电子信息工程学院,天津300401 [2]战略支援部队信息工程大学信息技术研究所,郑州450002 [3]天津市滨海新区信息技术创新中心,天津300450

出　　处：《计算机应用研究》2021年第7期2066-2071,共6页Application Research of Computers

基　　金：国家核高基重大专项基金资助项目。

摘　　要：针对拟态裁决器多数一致性表决算法的优化方法,提出用异常检测的方法直接量化数据可靠性来提升表决正确率。基于异常值的表决算法,通过构建拟态系统异构执行体输出数据集和训练深度学习异常检测模型量化了执行体输出数据异常值;使用权值优化算法优化加权分配,在表决时选择最优加权结果作为表决输出结果。实验结果表明,该方法能够提升拟态裁决器的表决输出正确率,具有一定共模逃逸检测能力,提升了系统的安全性和可靠性。This paper proposed an optimization method for majority consensus voting algorithm to improve the security of mimic adjudicators.This mimic ruling method based on executive outliers used anomaly detection to directly quantify data reliability to improve voting accuracy.Through constructing the executive body output data set and training the deep learning model,it quantified the abnormal value of the executive body output data.Then,it used the weight optimization algorithm to optimize the weighted distribution of the two parameters.Finally,it selected the optimal weighted result as the voting output result during voting.The experimental results show that the proposed method can improve the accuracy of mimic voting output and has a certain common mode escape detection capability.The optimization model can significantly improve the safety performance of the system.

关 键 词：拟态防御 裁决器 表决算法 深度学习 异常检测 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]