基于Double-Bagging特征降维异质集成入侵检测*陈俊彦,卢贤涛,黄雪锋,卢小烨,廖岑卉珊(桂林电子科技大学计算机与信息安全学院,广西桂林541004)摘要:入侵检测是网络安全领域中具有挑战性的重要任务。单个分类器可能会带来分类偏差,使用集成学习相较单分类器,具有更强的泛化能力及更高的精确率,但调整各基分类器的权重需要大量的时间。基于此问题,提出了一种基于Bagging特征降维和基于Bagging异质集成入侵检测分类算法(Double-Bagging)的特征降维异质集成入侵检测算法。该算法通过集成5个特征选择算法,采用Bagging投票机制选出最优特征子集,实现高效准确的特征降维。同时,引入集成学习中的成对多样性度量,从不同基分类器组合中选出最优异质集成集合。对于赋权函数综合使用精确率和AOC值作为权重对分类器进行集成。实验结果表明,所提算法精确率高达99.94%,系统错误率及正判率分别为0.03%和99.55%,均优于现有主流入侵检测算法的。关键词:入侵检测;异质集成学习;特征降维;成对多样性度量中图分类号:TP393文献标志码:Adoi:10.3969/j.issn.1007-130X.2023.06.008Double-BaggingbasedfeaturedimensionreductionheterogenousintegratedintrusiondetectionCHENJun-yan,LUXian-tao,HUANGXue-feng,LUXiao-ye,LIAO-CENHui-shan(SchoolofComputerScienceandInformationSecurity,GuilinUniversityofElectronicTechnology,Guilin541004,China)Abstract:Intrusiondetectionisachallengingandimportanttaskinthefieldofnetworksecurity.Asingleclassifiermaybringclassificationbias,andusingensemblelearninghasstrongergeneralizationabilityandhigheraccuracycomparedtoasingleclassifier.Althoughsuchalgorithmshavegoodclassifi-cationperformance,adjustingtheweightsbetweenthebaseclassifiersrequiresalotoftime.Toaddressthisissue,anfeaturedimensionreductionheterogenousintegrationintrusiondetectionmodelbasedonBagging-basedfeaturedimensionreductionandBaggingheterogeneousintegration-basedintrusiondetec-tionclassificationalgorithm(Double-Bagging)isproposed.Thealgorithmintegratesfivefeatureselec-tionalgorithmsandadoptsaBaggingvotingmechanismtoselecttheoptimalfeaturesubset,inordertoachieveefficientandaccuratefeaturedimensionalityreduction.Atthesametime,thepairwisediversitymeasureinensemblelearningisintroducedtochoose...
