国际会议演讲稿

自我介绍Thankyou,Mr./Ms.Chair./professorMynameissangqian.Iamveryhonoredtobeheretodooralpresentation.IamaMasterstudentfromHohaiUniversityandIamcurrentlydoingsomeresearchonphysicallayersecurity.Today,Iwouldliketosharewithyousomeofmyresearchonrelayselectionincooperativecommunication.(external/ek?st?rn?l;?k?st?rn?l)内容安排：Mypresentationincludesthesefiveparts.First,somebackgroundinformationaboutthisresearch;Second,systemmodelwehavedone;Third,NN-basedrelayselectionschemewehaveproposedForth,SimulationandresultsanalysisAndlast,someconclusionswehavegotP4Partone,introductionFirstly,Iwouldliketogiveyouabitofbackground.Differingfromthetraditionalcryptographictechniquesbasedonsecretkeys,wecanmakeuseofwirelesschannelcharacteristicstoenhancephysicallayersecurity.Cooperativecommunicationhasbeenwidelyrecognizedasaneffectivewaytocombatwirelessfadingandprovidediversitygainwhichisoneoftheresearchhotspots.Machinelearningasanemergingtechnologyhasbeenwidelyappliedinimageprocessing,cancerprediction,stockanalysisandotherfields.Sowhynottryitinwirelesscommunication?P5:Next,IwanttotalkalittlebitaboutpresentstudyRecentstudiesondeeplearningforwirelesscommunicationsystemshaveproposedalternativeapproachestoenhancecertainpartsoftheconventionalcommunicationsystemsuchasmodulationrecognition、channelencodinganddecodingchannelestimationanddetectionandanautoencoderwhichcanreplacethetotalsystemwithanovelarchitecture【modulationrecognition:AnNNarchitectureformodulationrecognitionthatconsistsofa4-layerNNandtwotwo-layerNNs。channelencodinganddecoding：AplainDNNarchitectureforchanneldecodingtodecodekbitsmessagesfromNbitsnoisycodewords。channelestimationanddetection:Adense-Netforsymbol-to-symboldetectioncanadoptlongshort-termmemory(LSTM)todetectanestimatedsymbol.Autoencoder:theautoencodercanrepresenttheentirecommunicationsystemandjointlyoptimizethetransmitterandreceiveroveranAWGNchannel.】P6Sowhydidweconductthisresearch?Well,wewanttoexploitthepotentialbenefitsofdeeplearninginenhancingphysicallayersecurityincooperative(?/k??'?p?r?t?v/?wirelesscommunicationandreducethefeedbackoverheadinlimitedspectrumresoucebyourourproposedschemeP8Nowletmemoveontoparttwo-systemmodelHere,youcanseeafigurewhichisasystemmodel.Thisisthesource;thesearetherelaynodesandthisisthedestination,thisistheeavesdropperThewholeprocessofcooperativewirelesscommunicationcanbedividedintotwophasesInthefirstphase,thesourcebroadcaststhesignaltotheoptimalrelaywhichguaranteesperfectsecurity.AsshowninFig1,hsr,irepresentsafadingcoefficientofthechannelfromthesourcetotherelaynode(R.)Inthesecondphase,theoptimalrelayforwardsascaledversionofitsreceivedsignaltothedestinationinthepresenceoftheeavesdropper,wheretheoptimalrelayisconsideredtoadoptamplify-and-forward(AF)relayscheme.Inthisfigure,hrd)representsafadingcoefficientofthechannelfromtherelayRtothedestinationgre,irepresentsafadingcoefficientofthechannelfromtherelayRtotheeavesdropper.P9:Hereyoucanseesomefollowingexpressions.Iamnotgoingtowasteourprecioustimeonthelengthyderivation.Iwouldliketoinviteyoutodirectlytakealookattheequationinitsfinalform.Thisistheoptimalindexoftheselectedrelaywiththeconventionalrelayselectionscheme.AmaongthisexpressionCs,i=max(Cd,i—Ce,i,0)representstheachievablesecrecyrateofsystemmodelwhenthei-threlayisselected.P11Nowletmemovetopartthree—NN-basedRelaySelectionHereyoucanseeafigurewhichshowsconventional3-layerneuralnetwork.Itconsistsofinputlayer,hiddenlayer1,hidden(/'h?dn/)layer2andoutputlayer.Neuralnetworkcanlearnfeaturesfromrawdataautomaticallyandadjustparameters(/p??r?m?t?(r)z/)flexibly^?/'fleks?bli/_suchasweightsandbiases.Incomplex(?/'k?mpleks/)conditions(scenarios(/s?'na?r???/),)Neuralnetworkhaspromisingapplicationsinrelayselectionforseveralreasons.First,thedeepnetworkhassuperior(/su??p??r??/learningabilitydespite(/d?'spa?t/)thecomplexchannelconditionsSecond,Neuralnetworkcanhandlelargedatasetsbecauseofdistributed(/d?'str?bj?t?d/)andparallel(/'p?r?lel/)computing(/k?m'pju?t??/s,whichensurecomputation(/k?mpj?'te??(?)n/(speedandprocessingcapacity(?/k?'p?s?t?).Third,variouslibrariesorframeworks,suchasTensorFlow,Theano,andCaffegiveitwideapplicationsInthispaper,theproblemoftherelayselectionismodeledasamulti(/'m?lt?/,ao)-classificationproblem.Weadoptsimpleneuralnetwork(NN)toselecttheoptimalrelaytoguaranteesperfectsecrecyperformaneeofrelaycooperativecommunicationsystem.(enhancephysicallayersecurity)P12Beforetrainingtheclassificationmodel,weneedtomakesomepreparationfordeeplearningtoacquireatrainingsetandatestingset.First,weneedtoproducerealfeaturevectorforeachexampleaccordingtochannelstateinformation;becausethechannelstateinformationmatricesiscomposedofcomplexnumbersbutfeaturevectorsaregenerallycomposedofrealnumbers.Soweneedtochangecomplexnumbersintorealnumberswithabsolute(/'?bs?lu?t/)valueoperation.Moreover,inordertoimprovetheclassificationperformance(precision),itisnecessarytonormalizethefeaturevectors.Second,weneedtodesignkeyperformaneeindicator(KPI).Inordertoeffectivelypreventtheeavesdropperfrominterceptinginformation,wechooseachievablesecrecyrateastheKPIofsystem.ThisKPIindicates(represents/shows)thediffereneeoftheachievableratefromthesourcetothedestinationandtheachievableratefromthesourcetotheeavesdropper.Third,wecanmakelabelsforexamplesaccordingtoKPI.theindexoftherelaywhichobtainsthemaximum(?/'m?ks?m?m/)KPIisregardedastheclasslabeloftheexample.P13ClassificationmodelThispicture(isabout)showsthewholeprocessofbuildingclassificationmodel.Thewholeprocessofbuildingclassificationcanbedividedintotwophases,namelytrainingphaseandtestingphase.Inthefirstphase,weneedtochoosesuitablehyper(?/'ha?p?/parameterstotrainneuralnetworkmodel.Inthesecondphase,wecanpredict(/pr?'d?kt/)labelsofoptimalrelayaccordingtoinputdataandassessclassificationperformanee.P15Nowletmemovetopartfour—SimulationandResults(?/r??z?ltsJAnalysisHere,youcanseeafigurewhichshowstherelationshipbetweentheaveragetransmit(?/tr?nz?m?t)powerofthesourceandtheachievablesecrecyratewithdifferentnumbersofrelays.Inthisfigure,thebluelinerepresentstheconventionalrelayselectionschemeandtheredlinerepresentstheNN-basedscheme.Inthisfigure,asthenumbersofrelayandtheaveragetransmitpowerofthesourceincreasestheachievablesecrecyrateincreasesaccordingly,.whichmeansincreasingthenumberofrelayscaneffectivelyimprovethesecretperformanee.Theredlinearealmost(?/'??lm??st)closetotheblueline,【whchindicatesthatourproposedscheme(i.e.theNN-basedscheme)achievesalmostthesamesecrecyratesasthoseoftheconventionalschemeforallvaluesofN,】whichvalidateseffectiveness(/?'fekt?vn?s/)ofourproposedschemes.P16Thistableshowsthethenormalized(/?n?rm??la?zd”meansquare(?/skwe?/)error(NMSES)valuesofdirrerentrelaynodes.ThevalueofNMSEmeanstheperformaneediffereneebetweentheconventionalschemeandourproposedscheme.ThevaluesofNMSEarebelow(/b?'l??/)negative,neg?t?v/)20(-20dB),whichvalidateseffectivenessofourproposedschemeagain.P17Now,letmemovetothelastpart—ConclusionOkay,nowwearegoingtotakealookatthelastpart-Conclusion.P18Wehavegotthefollowingconclusions.First,Incomplex(conditions)scenarios.Neuralnetworkhaspromisingapplicationsinrelayselectionforsuperiorlearningability,computationspeedandprocessingcapacity.Second,Comparedwiththeconventionalrelayselectionscheme,ourproposedschemeachievesalmostthesamesecrecyperformanee.Andlast,Ourproposedschemehasanadvantage(/?d'va?nt?d?/ofrelativelysmallfeedbackoverhead,indicatingthatproposedschemecanbeappliedtotheconditions(scenarios)wherethefeedbackislimited.(IftheconventionalschemeneedsfeedbackofNcomplexnumbers,NN-basedschemewillonlyneedfeedbackofNrealnumbers.Therefore,thefeedbackoverheadofourproposedschemeishalf(/ha??)ofthatoftheconventionalscheme,)Q&Ai计算复杂度ComputationalcomplexityThebiggestdrawbackisthehighlyselectioncomplexitieswithasmallnumberofrelaynodes.Ifnumberofrelaynodeisbig,itwillhaveaadvantage.Thisneedourfurtherresearch.Q:TheexperimentshowsthatsecrecyrateisalmostthesameastraditionalmethodandwhatisthepromotionofusingNNtorelayselection.(whatismeaningofintroducingNNtorelayselection)A:Thatourproposedscheme(i.e.theNN-basedscheme)achievesalmostthesameachievablesecrecyrateasthatoftheconventionalschemeindicatesthatourproposedschemeiseffectiveanditcanselectoptimalrelaynodewhichobtainsmaximumachievablesecrecyrate.Onereason(thefirstreason)isthatAdoptingNNforrelayselectionisanovelidea.Anotherreasonisthatthespectrumresourceisrelativelimitedandourproposedschemehassmallfeedbackoverhead.Q:what'sthemeaningof“perfectsecrecyperformance?What'sthemeaningof“Comparedtotheconventionalrelayselectionschenfe?A:“perfectsecrecyperformanee”meanstheachievablesecrecyrateisthebiggestonewhichcanenhancephysicallayersecurity.Infact,theconventionalrelayselectionschemeistheexhaustivesearch.Theindexofrelayselectionwiththisschemeisthebestone.Q：“Itis.obviosithatthe.feedback.overhead.ofproposed.scheme,.ishalf,of.that.of.theconventionalschemeA:well,Let'smakeanassumption.IftheconventionalschemeneedsfeedbackofNcomplexnumbers,NN-basedschemewillonlyneedfeedbackofNrealnumbers.Therefore,thefeedbackoverheadofourproposedschemeishalf(ha??)ofthatoftheconventionalschemeQ:1)Thetrainingdatasetandthetestingdatasetaresocalled"legitimatechannelcomplexmatrixandthewiretapchannelcomplexmatrix".2)Howthesedatacanbeobtained/generated?3)WhyitisimportantandsignificantastheinputoftheNN-basedapproachandwhatistheoutput?Thewholeprocessisnotclear.A:1)thedeeplearningneedstrainingsetandtestingset.trainingsetisusedtotrainmodelandtestingsetisusedtoassesthemodel'performanee.2)itcanbeobtainedaccordingtochannelstateinformationwherethelegitimatelinksandwiretaplinksaremodeledasRayleighfadingchannels.3)thechannelstateinformationisimportantfeature(indicator)inwirelessscommunication.thecommunicationtheNNcanlearnfeaturesfromrawdateA: