快速的包分類算法的研究.pdf

1、中南大學碩士學位論文快速的包分類算法的研究姓名：佘鋒申請學位級別：碩士專業(yè)：計算機應用技術指導教師：王小玲20090524ABSTRACTSincetheemergenceofInternet，ithassupportedthefunctionofeachdomesticfielddirectlyandplayedmoreandmoreimportantroleinthedevelopmentofnationaleconomyCurre

2、ntlyInternetgetsthegigabitorhighertransferspeedinthecommunicationnetworksHowever，thetraditionalroutershoulderingthecommunicationtask，whichoffersaBestofServiceandforwardspacketsundistinguishedlycan’tmeettherequirementofcu

3、stomizedservicesAndinordertoprovidedifferentiatedservicesforvarioususers，routersshouldclassifypacketsitreceivesbeforeforwardingTherefore，packetclassificationhasbecomethefoundationofdifferentiatedservicessuchasfirewallpac

4、ketfiltering，policybasedrouting，virtualprivatenetwork，trafficbillingandSOonHoweveratthesametime，ithastumedouttobeabottleneckofhighspeedrouterandbringsforwardtheproblemofefficientclassificationwithanacceptabletimeandspace

5、complexityInthisthesis，weproposeapriority—basedquad—tree(PQT)algorithmforpacketclassification，basedonrecursivespacedecompositionandthepriorityofrulesInconstructingaquadtreegeneratedbasedonrecursivespacedecomposition，thep

6、riorityofrulesisprimarilyconsideredintheproposedalgorithmInthesimulation，theproposedalgorithmachievesgoodperformanceintherequiredmemorysizeandreasonableperformanceintheclassificationspeedInthispaperweexploitdistinctivefl

7、owcharacteristicsofattackswhentheycommunicateonanetwork，andproposeasemisupervisedclassificationmethodthatcanaccommodatebothknownandunknownattacksIntrainingourclassifierweemploySFS(sequentialforwardselection)togetthebestf

8、eaturesubsetMeanwhile，weproposeweightedsamplingtechniquestoobtaintrainingflowsOurperformanceevaluationusingKDDCUP1999datashowsthathighflowandbyteclassificationaccuracyCanbeachievedKEYWORDSPacketclassification，quad—tree，r