基于模糊分類的入侵檢測.pdf

1、中南大學(xué)碩士學(xué)位論文基于模糊分類的入侵檢測姓名：唐曉衡申請學(xué)位級別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：夏利民20080601ABSTRACTWiththedevelopmentofcomputerandcommunicationtechnologydamagescausedbyunexpectedintrusionsandcrimesrelatedtocomputersystemshavebeenincreasingrapidlyThe

2、refore，networksecuritywhichCanensurethesystemtobehaveasintendedandtoprovidestableservicesbecomefocusIntrusionDetectionSystems(IDS)extractinformationfromacomputeroranetworkofcomputers，andattempttodetectthepresenceofintrus

3、ionsfromexternalsources，aswellassystemabusesbyauthorizedusersInordertowithstandmoreandmorefrequentcompoundnetworkattacksandhackercommitmentofdistribution，multiobjective，multistagenowadays，improveintrusiondetectionefficie

4、ncyunderthecircumstanceofhighbandwidthandlargescalenetwork，decreasefalsenegativerateandshortendetectiontime，itisnecessarytomakeimprovementonexistingintrusiondetectionmethodsTryingtoimprovetheperformanceoftheintrusiondete

5、ctionprocessfromalltheperspective，differentapproachesarepresentedinthispaperasfollows：(1)RedundantfeaturesCannotonlyoccupyhugestoragespaces，butalsoCandecreasetheaccuracyoftheclassifiersobeforethedetection。itisnecessaryto

6、analyzefeatureselectionforinputdataInthisthesis，theproposedalgorithmstOselectfeatureareintroducedTheinfluenceofargumentsonperformanceinparticleswarmoptimization(PSO)algorithmisanalyzedAnewalgorithmcombiningimmunesystemwi

7、thPSOisproposedtoeliminatetheredundancypropertyreducetheproblemsize，improvethequalityofclassificationandspeedupthedetectionThepositionoftheparticleisexpressedinabinarystring，theupdatestrategiesofthepositionandvelocityand

8、theselectionoffitnessfunctionaleillustratedindetailTheresultsshowthattheproposedalgorithmisefficientforfeatureselection(2)Inmanyyears，theresearchesonintrusiondetectionhavebeendevotedtoimprovethedetectionveracityrateandcu

9、tdownfalsealarmrateandmissingreportrateInthisthesis，thefuzzyclassificationisintroducedtointrusiondetectionFuzzyrulesinvolvedinintrusiondetectionareobtainedbyGeneticalgorithm；andBoostingalgorithmisemployedtochangethedistr

10、ibutionoftraininginstancesduringeachroundoftraining，SOthatthefuzzyclassificationrulenewlyextractedbyGeneticalgorithmwillputmoreemphasisupontheinstancesmisclassifiedoruncoveredAndweightedvotingmethodisusedtointegratethefu