摘要
针对机载网络入侵检测系统样本具有不平衡性的问题,提出多元状态估计技术和基于分解的多目标进化算法(MSET-MOEA/D)相结合的方法,实现异常样本极少的航空数据总线异常检测。研究基于数据总线观测样本和估计样本偏差的初始预警阈值MSET评估算法,提出一种综合考虑误报数和漏报数的阈值设置方法。该方法以裕量为自变量,以误报数和漏报数为目标,将异常检测问题转变为一个多目标优化问题。进一步提出基于MOEA/D的可变目标权重的Pareto解集评价方法,获得最高满意度对应的折中解。机载网络数据集以GJB289航空总线数据集开展算法的验证和比较研究。实验结果表明,所提算法不仅适用于可变的多目标权重要求,且具有漏报数或误报数更低、检测效率高的优势,实用价值更高。
Aiming at the problem that the samples of the airborne network intrusion detection system are unbalanced,a method combining the multi-state estimation technology and the multi-objective evolutionary algorithm based on decomposition(MSET-MOEA/D)is proposed to achieve the anomaly detection of the aviation data bus with very few abnormal samples.The research focuses on the MSET evaluation algorithm of the initial early warning threshold based on the deviation of the observed samples and estimated samples of the data bus,and proposes a threshold setting method that comprehensively considers the number of false alarms and missed alarms.This method takes the margin as the independent variable and the number of false alarms and missed alarms as the goals,transforming the anomaly detection problem into a multi-objective optimization problem.Furthermore,a Pareto solution set evaluation method with variable target weights based on MOEA/D is proposed to obtain the compromise solution corresponding to the highest satisfaction.In this paper,the airborne network dataset conducts the verification and comparative study of the algorithm using the GJB289 aviation bus dataset.The experimental results show that the proposed algorithm is not only applicable to the variable multi-objective weight requirements,but also has lower missed or false alarms,faster detection,and higher practical value.
作者
闫洁
王晓前
刘文琪
葛红娟
YAN Jie;WANG Xiao-qian;LIU Wen-qi;GE Hong-juan(The 54th Research Institute of China Electronics Technology Group Corporation,Shijiazhuang 050000,China;Avic China Helicopter Research and Development Institute,Jingdezhen 333000,China;Nanjing University of Aeronautics and Astronautics,Nanjing 211000,China)
出处
《航空计算技术》
2026年第1期53-58,共6页
Aeronautical Computing Technique
基金
国家自然科学基金与民航基金联合重点资助项目(U2133203,U2233205)。
