資源描述:
《基于進(jìn)化半監(jiān)督式模糊聚類算法的入侵檢測(cè)》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、西安電子科技大學(xué)碩士學(xué)位論文基于進(jìn)化半監(jiān)督式模糊聚類算法的入侵檢測(cè)姓名:王曉峰申請(qǐng)學(xué)位級(jí)別:碩士專業(yè):計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師:姜建國20050101摘要入鏝撿{夔
2、
3、技術(shù)莛繼“防火壤”、“數(shù)豢熱密”等簧緩安全縑妒設(shè)蘸囂瓣瑟一代安全保障技術(shù)。它可以乇H別針對(duì)計(jì)算機(jī)葳網(wǎng)絡(luò)資源的惡意企圖和行為,并對(duì)此作出甥應(yīng)的反應(yīng),已越來越成為絳障網(wǎng)絡(luò)安全豹羹要手段之一。本文討論了入侵槍測(cè)技術(shù)的分類、常用檢測(cè)方法及工作原理,分析了目前入侵撿測(cè)漿繞(IDS)存在鮑主要問題秘發(fā)鼴趨勢(shì)。糖崮蘩捻濺算法瓣疆究楚IDS瓣主要發(fā)展方向之一。本文將進(jìn)化半監(jiān)
4、督式模糊羧糞(ESSFC)算法應(yīng)魘到IDS中??K惴ㄊ谷成凫嗫顦?biāo)簽數(shù)掘和大量未知標(biāo)簽數(shù)據(jù)生成IDS的分?jǐn)\器。非常適合予標(biāo)簽數(shù)掘較難獲得而未知標(biāo)簽數(shù)據(jù)容易獲褥的蛾?duì)睢SSFC算法可處理模糊類標(biāo)簽,不茹陷入弱鄭最優(yōu),適合于黼糟優(yōu)化表面或擁有多個(gè)局部?jī)?yōu)化解時(shí)凝值函數(shù)的優(yōu)化。ESSFC算法能勢(shì)行操作,適合并行結(jié)構(gòu)的實(shí)現(xiàn)。用ESSFC算法生成的分類器對(duì)入侵?jǐn)?shù)搬送行分類,操作簡(jiǎn)單,運(yùn)算量小,檢測(cè)率高,非常適合予實(shí)際成用。實(shí)驗(yàn)結(jié)果裝明,相對(duì)于基于模糊C均值聚嫫算法的IDS,基于ESSFC算法的tDS裔較高的檢測(cè)率。相對(duì)于基于支持向量
5、視算法的1DS有對(duì)標(biāo)簽數(shù)據(jù)需求少的優(yōu)勢(shì)。綜合考慮,基予ESSFC算法的IDS是一神較好的IDS。關(guān)鍵詞;入侵檢測(cè)網(wǎng)絡(luò)蜜全半監(jiān)督學(xué)習(xí)進(jìn)化規(guī)劃聚類2基于、
6、,監(jiān)磐式模糊聚類莓法的入侵檢測(cè)ABSTltACTTheIntmsionDetectiontechnologyisanewsecuritytechnology,apartfromtraditionalsecurityprotectingtechnology’suchasfirewallanddataencryption.Itmonitorsthecomputerornetw
7、orkandreactslotheviciousintrusionorsuspiciousactivities.Ithasbecomemoreandmoreimportant.TheclassoftheIntrusionDetectiontechnology,commondetectionmethodandworktheorywasdiscussedfirstly.ThenthemainproblemsofthepresentIDSandthedevelopingtrendwereanalyzed,indicatingt
8、hatthestudyofnewdetectionalgorithmisoneofmaindevelopingaim.Thispaperproposedamethodthatapplysanevolutionarysemi—supervisedfuzzyclustering(ESSFC)algorithmtointrusiondetectionsystem(IDS).ThisalgorithmrequiresasmallnumberofIabeleddataonlyandalargenumberofunlabeledda
9、ta.ItiSwellsuitedforthestatusquothatgainedlabeleddatadifficultlyandgmnedunlabeleddataeasybysomeinstruments.Thisalgoritluncandealwithfuzzylabelanduneasilygetintolocallyoptima.Real—valuedfunctionoptimization,inwhichtheoptimizationsurfaceis“rugged”orprocessesmanyloc
10、allyoptima,andiswellsuitedforESSFCalgorithm,Moreover,ESSFCalgorithmisabletOexhibitparallelism,andissuitedtoimplementonmassivelyparallelarchitecture.TheclassifierproducedbyESSFCalgorithmclassifiestheintrusiondata,whichiseasytohandle,calculatesafewandhashi曲detectio
11、nefficiency.Therefore,itiswellsuitedforapplication.ExperimentalresultsindicatethattheIDSbasedonESSFCalgorithmcarlimproveclassificationaccuracymoresignificantly