Self Learning Network Traffic Classification

《Self Learning Network Traffic Classification 》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、IEEESponsored2ndInternationalConferenceonInnovationsinInformationEmbeddedandCommunicationSystemsICIIECS’15SelfLearningNetworkTrafficClassificationVandanaM,SruthyManmadhan,Student,M.TechAssistantProfessorComputerScience&Engineering,ComputerScience&EngineeringNSSCollegeofEngineering,Palakk

2、adKeralaNSSCollegeofEngineering,Palakkad,vandana19191@gmail.comKeralasruthym.88@gmail.comAbstract-Networkmanagementispartoftrafficengineeringthenetworkmanagersmusthaveadetailedknowledgeaboutandsecurity.Thecurrentsolutions-DeepPacketInspectionapplicationsandprotocols.Theuserapplicationmay

3、allow(DPI)andstatisticalclassification,relyontheavailabilityofalargedelaysorjitter,buttheusersmightbeverysensitivetotrainingset.Incaseofthesethereisacumbersomeneedtolongwaittimes.Managingnetworktrafficrequiresaregularlyupdatethesignatures.Furthertheirvisibilityisjudiciousbalanceofallthes

4、epriorities.limitedtoclassestheclassifierhasbeentrainedfor.UnsupervisedalgorithmshavebeenenvisionedasaanClassificationoftraffichelpsidentifydifferentalternativetoautomaticallyidentifyclassesoftraffic.applicationsandprotocolsthatexistinanetwork.DifferentToaddresstheseissuesSelfLearningNet

5、workmethodssuchasmonitoring,discovery,control,andTrafficClassificationisproposed.Itusesunsupervisedoptimizationcanbeperformedontheidentifiedtrafficinalgorithmsalongwithanadaptiveseedingapproachtoordertoimprovethenetworkperformance.Typically,onceautomaticallyletsclassesoftraffictoemerge,m

6、akingthemthepacketsareclassified(identified)asbelongingtoaidentifiedandlabelled.Unliketraditionalclassifiers,thereisnoparticularapplicationorprotocol,theyaremarkedorflagged.needofa-prioriknowledgeofsignaturesnoratrainingsettoextractthesignatures.Instead,SelfLearningNetworkTrafficThesemar

7、kingsorflagshelptherouterdetermineappropriateClassificationautomaticallygroupsflowsintopure(orservicepoliciestobeappliedforthoseflows.homogeneous)clustersusingsimplestatisticalfeatures.ThisAllgenericclassificationtechniquesbasedonlabelassignment(whichisstillbasedonsomeman