通信类英文文献及翻译.docx
《通信类英文文献及翻译.docx》由会员分享,可在线阅读,更多相关《通信类英文文献及翻译.docx(11页珍藏版)》请在冰豆网上搜索。
通信类英文文献及翻译
姓名:
刘峻霖班级:
通信143班学号:
2014101108
附录
一、英文原文:
DetectingAnomalyTrafficusingFlowDataintherealVoIPnetwork
I.INTRODUCTION
Recently,manySIP[3]/RTP[4]-basedVoIPapplicationsandserviceshaveappearedandtheirpenetrationratioisgraduallyincreasingduetothefreeorcheapcallchargeandtheeasysubscriptionmethod.Thus,someofthesubscriberstothePSTNservicetendtochangetheirhometelephoneservicestoVoIPproducts.Forexample,companiesinKoreasuchasLGDacom,SamsungNet-works,andKThavebeguntodeploySIP/RTP-basedVoIPservices.ItisreportedthatmorethanfivemillionusershavesubscribedthecommercialVoIPservicesand50%ofalltheusersarejoinedin2009inKorea[1].AccordingtoIDC,itisexpectedthatthenumberofVoIPusersinUSwillincreaseto27millionsin2009[2].Hence,astheVoIPservicebecomespopular,itisnotsurprisingthatalotofVoIPanomalytraffichasbeenalreadyknown[5].So,MostcommercialservicesuchasVoIPservicesshouldprovideessentialsecurityfunctionsregardingprivacy,authentication,integrityandnon-repudiationforpreventingmalicioustraffic.Particu-larly,mostofcurrentSIP/RTP-basedVoIPservicessupplytheminimalsecurityfunctionrelatedwithauthentication.Thoughsecuretransport-layerprotocolssuchasTransportLayerSecurity(TLS)[6]orSecureRTP(SRTP)[7]havebeenstandardized,theyhavenotbeenfullyimplementedanddeployedincurrentVoIPapplicationsbecauseoftheoverheadsofimplementationandperformance.Thus,un-encryptedVoIPpacketscouldbeeasilysniffedandforged,especiallyinwirelessLANs.Inspiteofauthentication,theauthenticationkeyssuchasMD5intheSIPheadercouldbemaliciouslyexploited,becauseSIPisatext-basedprotocolandunencryptedSIPpacketsareeasilydecoded.Therefore,VoIPservicesareveryvulnerabletoattacksexploitingSIPandRTP.WeaimatproposingaVoIPanomalytrafficdetectionmethodusingtheflow-basedtrafficmeasurementarchi-tecture.WeconsiderthreerepresentativeVoIPanomaliescalledCANCEL,BYEDenialofService(DoS)andRTPfloodingattacksinthispaper,becausewefoundthatmalicioususersinwirelessLANcouldeasilyperformtheseattacksintherealVoIPnetwork.FormonitoringVoIPpackets,weemploytheIETFIPFlowInformationeXport(IPFIX)[9]standardthatisbasedonNetFlowv9.Thistrafficmeasurementmethodprovidesaflexibleandextensibletemplatestructureforvariousprotocols,whichisusefulforobservingSIP/RTPflows[10].InordertocaptureandexportVoIPpacketsintoIPFIXflows,wedefinetwoadditionalIPFIXtemplatesforSIPandRTPflows.Furthermore,weaddfourIPFIXfieldstoobserve802.11packetswhicharenecessarytodetectVoIPsourcespoofingattacksinWLANs.
II.RELATEDWORK
[8]proposedafloodingdetectionmethodbytheHellingerDistance(HD)concept.In[8],theyhavepre-sentedINVITE,SYNandRTPfloodingdetectionmeth-ods.TheHDisthedifferencevaluebetweenatrainingdatasetandatestingdataset.ThetrainingdatasetcollectedtrafficovernsamplingperiodofdurationΔt.Thetestingdatasetcollectedtrafficnextthetrainingdatasetinthesameperiod.IftheHDiscloseto‘1’,thistestingdatasetisregardedasanomalytraffic.Forusingthismethod,theyassumedthatinitialtrainingdatasetdidnothaveanyanomalytraffic.Sincethismethodwasbasedonpacketcounts,itmightnoteasilyextendedtodetectotheranomalytrafficexceptflooding.Ontheotherhand,[11]hasproposedaVoIPanomalytrafficdetectionmethodusingExtendedFiniteStateMachine(EFSM).[11]hassuggestedINVITEflooding,BYEDoSanomalytrafficandmediaspammingdetectionmethods.However,thestatemachinerequiredmorememorybecauseithadtomaintaineachflow.[13]haspresentedNetFlow-basedVoIPanomalydetectionmethodsforINVITE,REGIS-TER,RTPflooding,andREGISTER/INVITEscan.How-ever,theVoIPDoSattacksconsideredinthispaperwerenotconsidered.In[14],anIDSapproachtodetectSIPanomalieswasdeveloped,butonlysimulationresultsarepresented.FormonitoringVoIPtraffic,SIPFIX[10]hasbeenproposedasanIPFIXextension.ThekeyideasoftheSIPFIXareapplication-layerinspectionandSDPanalysisforcarryingmediasessioninformation.Yet,thispaperpresentsonlythepossibilityofapplyingSIPFIXtoDoSanomalytrafficdetectionandprevention.WedescribedthepreliminaryideaofdetectingVoIPanomalytrafficin[15].ThispaperelaboratesBYEDoSanomalytrafficandRTPfloodinganomalytrafficdetec-tionmethodbasedonIPFIX.Basedon[15],wehaveconsideredSIPandRTPanomalytrafficgeneratedinwirelessLAN.Inthiscase,itispossibletogeneratethesimiliaranomalytrafficwithnormalVoIPtraffic,becauseattackerscaneasilyextractnormaluserinformationfromunencryptedVoIPpackets.Inthispaper,wehaveextendedtheideawithadditionalSIPdetectionmethodsusinginformationofwirelessLANpackets.Furthermore,wehaveshowntherealexperimentresultsatthecommercialVoIPnetwork.
III.THEVOIPANOMALYTRAFFICDETECTIONMETHOD
A.CANCELDoSAnomalyTrafficDetection
AstheSIPINVITEmessageisnotusuallyencrypted,attackerscouldextractfieldsnecessarytoreproducetheforgedSIPCANCELmessagebysniffingSIPINVITEpackets,especiallyinwirelessLANs.Thus,wecannottellthedifferencebetweenthenormalSIPCANCELmessageandthereplicatedone,becausethefakedCANCELpacketincludesthenormalfieldsinferredfromtheSIPINVITEmessage.TheattackerwillperformtheSIPCANCELDoSattackatthesamewirelessLAN,becausethepurposeoftheSIPCANCELattackistopreventthenormalcallestab-lishmentwhenavictimiswaitingforcalls.Therefore,assoonastheattackercatchesacallinvitationmessageforavictim,itwillsendaSIPCANCELmessage,whichmakesthecallestablishmentfailed.WehavegeneratedfakedSIPCANCELmessageusingsniffedaSIPINVITEmessage.FieldsinSIPheaderofthisCANCELmessageisthesameasnormalSIPCANCELmessage,becausetheattackercanobtaintheSIPheaderfieldfromunencryptednormalSIPmessageinwirelessLANenvironment.ThereforeitisimpossibletodetecttheCANCELDoSanomalytrafficusingSIPheaders,weusethedifferentvaluesofthewirelessLANframe.Thatis,thesequencenumberinthe802.11framewilltellthedifferencebetweenavictimhostandanattacker.WelookintosourceMACaddressandsequencenumberinthe802.11MACframeincludingaSIPCANCELmessageasshowninAlgorithm1.WecomparethesourceMACaddressofSIPCANCELpacketswiththatofthepreviouslysavedSIPINVITEflow.IfthesourceMACaddressofaSIPCANCELflowischanged,itwillbehighlyprobablethattheCANCELpacketisgeneratedbyaunknownuser.However,thesourceMACaddresscouldbespoofed.Regarding802.11sourcespoofingdetection,weemploythemethodin[12]thatusessequencenumbersof802.11frames.Wecalculatethegapbetweenn-thand(n-1)-th802.11frames.Asthesequencenumberfieldina802.11MACheaderuses12bits,itvariesfrom0to4095.WhenwefindthatthesequencenumbergapbetweenasingleSIPflowisgreaterthanthethresholdvalueofNthatwillbesetfromtheexperiments,wedeterminethattheSIPhostaddressasbeenspoofedfortheanomalytraffic.
B.BYEDoSAnomalyTrafficDetection
IncommercialVoIPapplications,SIPBYEmessagesusethesameauthenticationfieldisincludedintheSIPIN-VITEmessageforsecurityandaccountingpurposes.How-ever,attackerscanreproduceBYEDoSpacketsthroughsniffingnormalSIPINVITEpacketsinwirelessLANs.ThefakedSIPBYEmessageissamewiththenormalSIPBYE.Therefore,itisdifficulttodetecttheBYEDoSanomalytrafficusingonlySIPheaderinformation.AftersniffingSIPINVITEmessage,theattackeratthesameordifferentsubnetscouldterminatethenormalin-progresscall,becauseitcouldsucceedingeneratingaBYEmessagetotheSIPproxyserver.IntheSIPBYEattack,itisdifficulttodistinguishfromthenormalcallterminationprocedure.Thatis,weapplythetimestampofRTPtrafficfordetectingtheSIPBYEattack.Generally,afternormalcalltermination,thebi-directionalRTPflowisterminatedinabrefspaceoftime.However,ifthecallterminationprocedureisanomaly,wecanobservethatadirectionalRTPmediaflowisstillongoing,whereasanattackeddirectionalRTPflowisbroken.Therefore,inordertodetecttheSIPBYEattack,wedecidethatwewatchadirectionalRTPflowforalongtimethresholdofNsecafterSIPBYEmessage.ThethresholdofNisalsosetfromtheexperiments.Algorithm2explainstheproceduretodetectBYEDoSanomaltrafficusingcapturedtimestampoftheRTPpacket.WemaintainSIPsessioninformationbetweenclientswithINVITEandOKmessagesincludingthesameCall-IDand4-tuple(source/destinationIPAddressandportnumber)oftheBYEpacket.WesetatimethresholdvaluebyaddingNsectothetimestampvalueoftheBYEmessage.ThereasonwhyweusethecapturedtimestampisthatafewRTPpacketsareobservedunder0.5second.IfRTPtrafficisobservedafterthetimethreshold,thiswillbeconsideredasaBYEDoSattack,becausetheVoIPsessionwillbeterminatedwithnormalBYEmessages.C.RTPAnomalyTrafficDetectionAlgorithm3describesanRTPfloodingdetectionmethodthatusesSSRCandsequencenumbersoftheRTPheader.DuringasingleRTPsession,typically,thesameSSRCvalueismaintained.IfSSRCischanged,itishighlyprobablethatanomalyhasoccurred.Inaddition,ifthereisabigsequencenumbergapbetweenRTPpackets,wedeterminethatanomalyRTPtraffichashappened.Asinspectingeverysequencenumberforapacketisdifficult,wecalculatethesequencenumbergapusingthefirst,last,maximumandminimumsequencenumb