基于粒子群优化支持向量机的异常入侵检测-研究

《基于粒子群优化支持向量机的异常入侵检测-研究》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、硕士学位论文基于粒子群优化支持向量机的异常入侵检测研究过仿真表明，粒子群优化算法对于选取支持向量机参数是一种的有效方法，可以取得令人满意的效果。关键词：入侵检测；异常检测；粒子群优化算法；支持向量机；参数选择硕士学位论文基于粒子群优化支持向量机的异常入侵检测研究AbstractWiththefastdevelopmentofcomputernetworktechnology,thetrendistocommunicategloballyusingcomprehensiveopennetworkenvironment．Thenetworkprovidestheopenandshar

2、edresources，butthereisalwayssecurityrisk．Thefirewall，oncethemostpopulardefensivemethod，cannolongermeetpeople’Sdemandofnetworksecurity,theusersofnetworkconfrontwiththegraduallygravesafeproblemandnetwork’Sinvasionhasbecomethemostterriblethreatensofcomputerandnetwork’Ssafe．Asanimportantandactive

3、securitymechanism，IntrusionDetection(ID)willreinforcethetraditionalsystemsecuritymechanism．Intrusiondetectionplaysmoreimportantroleinnetworksecuritytoday．Thispaperintroducesamethod，particleswarmoptimizationandsupportvectormachine，tointrusiondetectionsystem，andpresentsanewdesignofIDbasedonPart

4、icleSwarmOptimization(PSO)andSupportVectorMachine(SVM)．Supportvectormachine，asanewkindofsoftsensortechniques，hasbeenstudiedwidelyintheworldrecently．SupportvectormachineisbasedonVapnik’Sminimalofthestructurerisk，triesitsbesttoincreasethegeneralization．Whenusingthemethodofsupportvectormachinein

5、tointrusiondetectionsystem，betterclassificationcanbeacquiredattheconditionthatthereislessknownknowledge．Sothemethodisappliedintheintrusiondetectionsystem．Thesupportvectormachineparameterdecidesitsstudyperformanceandexudestheability．Astheparameterchoiceisinfinite，theparameterchoiceneedsenormou

6、stime，andisverydifficulttoapproachsuperiorly．SincetheSVMmodeldependonapropersettingofitsparameters(regulationparameterCandtheradialbasisfunctionwidthparameter盯)，especiallyontheinteractionofthetwoparameters，thispaperpresentsanoptimalselectionapproachoftheSVMparametersbasedonparticleswarmoptimi

7、zationalgorithm．PSOisanewbiologicalevolutionaryalgorithm，originationfromthebehaviorstudyofbirds’seekingfood．Itcallbeimplementedwitheasyprinciplesandafewparametersneedtobetuned，aswellasithasmaximumstrengthindealingwithhigh—dimensionoptimizatio