SECUREANDEFFICIENTSPECTRUMSHARINGANDQOSANALYSISINOFDM-BASEDHETEROGENEOUSWIRELESSNETWORKSByAhmedS.AlahmadiADISSERTATIONSubmittedtoMichiganStateUniversityinpartialentoftherequirementsforthedegreeofElectricalEngineering|DoctorofPhilosophy2016ABSTRACTSECUREANDEFFICIENTSPECTRUMSHARINGANDQOSANALYSISINOFDM-BASEDHETEROGENEOUSWIRELESSNETWORKSByAhmedS.AlahmadiTheInternetofThings(IoT),whichnetworksversatiledevicesforinformationexchange,remotesensing,monitoringandcontrol,isingpromisingapplicationsinnearlyeveryHowever,duetoitshighdensityandenormousspectrumrequirement,thepracticaldevelopmentofIoTtechnologyseemstobenotavailableuntilthereleaseofthelargemil-limeterwave(mmWave)band(30GHz-300GHz).Comparedtoexistinglowerbandsystems(suchas3G,4G),mmWavebandsignalsgenerallyrequirelineofsight(LOS)pathandfromseverefadingleadingtomuchsmallercoveragearea.Fornetworkdesignandmanagement,thisimpliesthat:(i)MmWavebandalonecouldnotsupporttheIoTnetworks,buthastobeintegratedwiththeexistinglowerbandsystemsthroughsecureandespectrumsharing,especiallyinthelowerfrequencybands;and(ii)TheIoTnetworkswillhaveveryhighdensitynodedistribution,whichisatchallengeinnetworkdesign,especiallywiththescarceenergybudgetofIoTapplications.Motivatedbytheseobservations,inthisdissertation,weconsiderthreeproblems:(1)Howtoachievesecureandespectrumsharing?(2)HowtoaccommodatetheenergylimitedIoTdevices?(3)HowtoevaluatetheQualityofService(QoS)inthehighdensityIoTnetworks?Weaimtodevelopinnovativetechniquesforthedesign,evaluationandmanagementoffutureIoTnetworksunderbothbenignandhostileenvironments.Themaincontributionsofthisdissertationareoutlinedasfollows.First,wedevelopasecureandtspectrumsharingschemeinsingle-carrierwirelessnetworks.Cognitiveradio(CR)isakeyenablingtechnologyforspectrumsharing,wheretheunoccupiedspectrumisidenforsecondaryusers(SUs),withoutinterferingwiththeprimaryuser(PU).AserioussecuritythreattotheCRnetworksisreferredtoasprimaryuseremulationattack(PUEA),inwhichamalicioususer(MU)emulatesthesignalcharacteristicsofthePU,therebycausingtheSUstoerroneouslyidentifytheattackerasthePU.Here,weconsiderfull-bandPUEAdetectionandproposeareliableAES-assistedDTVscheme,whereanAES-encryptedreferencesignalisgeneratedattheDTVtransmitterandusedasthesyncbitsoftheDTVdataframes.ForPUdetection,weinvestigatethecross-correlationbetweenthereceivedsequenceandreferencesequence.TheMUdetectioncanbeperformedbyinvestigatingtheauto-correlationofthereceivedsequence.Wefurtherdevelopasecureandtspectrumsharingschemeinmulti-carrierwirelessnetworks.Weconsidersub-bandmalicioususerdetectionandproposeasecureAES-basedDTVscheme,wheretheexistingreferencesequenceusedtogeneratethepilotsymbolsintheDVB-T2framesisencryptedusingtheAESalgorithm.TheresultedsequenceisexploitedforaccuratedetectionoftheauthorizedPUandtheMU.Second,wedevelopanenergytransmissionschemeinCRnetworksusingenergyharvesting.WeproposeatransmittingschemefortheSUssuchthateachSUcanperforminformationreceptionandenergyharvestingsimultaneously.Weperformsum-rateopti-mizationfortheSUsunderPUEA.Itisobservedthatthesum-rateoftheSUnetworkcanbeimprovedtlywiththeenergyharvestingtechnique.Potentially,theproposedschemecanbeapplieddirectlytotheenergy-constrainedIoTnetworks.Finally,weinvestigateQoSperformanceanalysismethodologies,whichcanprovidein-sightfulfeedbackstoIoTnetworkdesignandplanning.TakingthespatialrandomnessoftheIoTnetworkintoconsideration,weinvestigatecoverageprobability(CP)andblock-ingprobability(BP)inrelay-assistedOFDMAnetworksusingstochasticgeometry.Moresp,wemodeltheinter-cellinterferencefromtheneighboringcellsateachtypicalnode,andderivetheCPinthedownlinktransmissions.Basedontheirdataraterequire-ments,weclassifytheincomingusersintotclasses,andcalculatetheBPusingthemulti-dimensionallossmodel.CopyrightbyAHMEDS.ALAHMADI2016Thisdissertationisdedicatedtomybelovedfamily...vACKNOWLEDGMENTSIwouldliketotakethisopportunitytoexpressmysincereappreciationforallthesupportandencouragementthathaveledtothecompletionofthisdissertation.Iamgreatlyin-debtedtomyadvisor,Prof.TongtongLi,forhercontinuoussupport,help,patience,andencouragementthroughoutmyPhDstudiesatMichiganStateUniversity.IwouldalsoliketothankProf.HassanKhalil,Prof.EricTorng,andProf.MiZhangforservingonmycommittee.Iamdeeplygratefultothemfortheirvaluablecommentsandinsightfuldiscussions.SpecialthanksgotomypastandpresentcolleaguesintheBroadbandAccessandWirelessCommunication(BAWC)lab:ZhaoxiFang,MaiAbdelhakim,TianlongSong,ZheWang,YuanLiang,RunTian,andYuZhengformakingmystayatMichiganStateUniversitymoreenjoyable,andfortheirinsightfulandhelpfulacademicinputs.Iamgreatlygratefultomyfamilyandfriendsfortheirtremendoussupportanden-couragement.Myprofoundgratitudeistomymotherforherendlesslove,prayers,andsupport.Icouldnotbemoregratefultomybelovedwifeandmysweetdaughter.Thisdissertationwouldnothavebeenaccomplishedwithouttheirlove,care,patience,andsupport.Thankyouallformakingthisdreamareality.viTABLEOFCONTENTSLISTOFTABLES....................................xLISTOFFIGURES...................................xiKEYTOABBREVIATIONS.............................xiiiChapter1Introduction...............................11.1Overview......................................11.1.1MotivationsandProblemIdenn.................11.1.2ResearchOverview............................21.2RevisitoftheExistingWork...........................41.2.1PrimaryUserEmulationAttack.....................41.2.2EnergyHarvestingandReliableInformationTransfer.........71.2.3QoSPerformanceMeasureinIoTNetworks..............91.3SummaryofDissertationContributions.....................10Chapter2Full-bandPUEADetectionandMitigation............152.1Introduction....................................162.2TheProposedAES-assistedDTVApproach..................182.2.1ABriefReviewoftheTerrestrialDTVSystem.............182.2.2AES-assistedDTVTransmitter.....................192.2.3AES-assistedDTVReceiver.......................212.2.3.1DetectionofthePrimaryUser.................222.2.3.2DetectionoftheMaliciousUser................232.2.4FurtherDiscussions............................252.3AnalyticalEvaluationoftheProposedAES-assistedDTVApproach.....272.3.1AnalyticalEvaluationofPrimaryUserDetection...........272.3.2AnalyticalEvaluationofMaliciousUserDetection...........312.3.2.1FalseAlarmRateandMissDetectionProbabilityforMali-ciousUserDetection......................312.3.2.2TheOptimalThresholdsforMaliciousUserDetection...382.4SecurityandFeasibilityoftheProposedAES-assistedDTVApproach....402.4.1ABriefOverviewoftheAESAlgorithm................402.4.2SecurityoftheAES-assistedDTV....................422.4.3Feasibility.................................432.5SimulationResults................................442.6Summary.....................................47Chapter3Sub-bandPUEADetectionandMitigation............493.1Introduction....................................49vii3.2SystemModel...................................513.3AES-basedPUEADetectionScheme......................543.3.1PrimaryUserTransmitterDesign....................543.3.2SecondaryUserCoordinatorReceiverDesign..............553.3.2.1PrimaryUserDetection....................563.3.2.2MaliciousUserDetection...................573.3.3FalseAlarmRateandMissDetectionProbability...........593.3.4DetectionPerformanceVersusSampleSize...............603.4SimulationResults................................613.5Summary.....................................62Chapter4EnergyHarvestingandReliableInformationTransfer.....644.1Introduction....................................654.2TheProposedEnergyHarvestingScheme....................664.3EnergyHarvestingandPerformanceOptimization...............684.3.1SecondaryUsersEnergyHarvestingandInformationRecovery....684.3.1.1AchievableTransmissionRateforSU1............694.3.1.2AchievableTransmissionRateforSU2............714.3.2Sum-rateOptimization..........................724.3.3SuboptimalSolution...........................744.3.3.1Sum-rateLowerBound.....................744.3.3.2SuboptimalSolutions......................774.4DiscussionsonWorstJammingInterferenceinPUEA.............784.5SimulationResults................................814.6Summary.....................................84Chapter5BlockingProbabilityAnalysisforRelay-assistedOFDMANet-works....................................865.1Introduction....................................865.2SystemModel...................................895.2.1NetworkModel..............................895.2.2TransmissionProtocolDesign......................905.2.3ChannelModel..............................925.3InterferenceModelingandAchievableRateAnalysis..............925.3.1InterferenceModel............................925.3.2CoverageProbability...........................945.3.2.1CoverageProbabilityofDirectTransmission.........945.3.2.2CoverageProbabilityof2-hopTransmission.........955.3.3AchievableTransmissionRate......................965.4BlockingProbabilityAnalysis..........................975.4.1ResourceAllocationandTModeling................985.4.2BlockingProbabilityCalculation....................1005.5SimulationResults................................1025.6Summary.....................................104viiiChapter6ConclusionsandFutureWork....................1066.1Conclusions....................................1066.2FutureWork....................................108APPENDIX........................................110BIBLIOGRAPHY....................................114ixLISTOFTABLESTable4.1:SystemParameters.............................82Table5.1:SimulationParameters...........................102xLISTOFFIGURESFigure1.1:ApossiblescenariofortheattackerstoavoidPUEAdetectionapproachesbasedonthelocationand/ortheenergylevelofthereceivedsignal.Forexample,MU1canproducethesameDOAandcomparablereceivedpowerlevelastheprimaryuser,whileMU2canproducecomparablereceivedpowerlevelastheprimaryuser..................6Figure2.1:8-VSBsignalframestructure........................19Figure2.2:Generationofthereferencesignal......................20Figure2.3:AESencryption................................41Figure2.4:Normalizedcross-correlationbetweenthereferencesignalandnoisyver-sionsofmalicioususer'ssignal.Notethatthecross-correlationvaluesareintheorderof104,whichiscloseto0................42Figure2.5:Normalizedcross-correlationbetweenthereferencesignalandnoisyver-sionsoftheprimaryuser'ssignal.Here,˙2s=1..............43Figure2.6:Example1:Thefalsealarmrateandmissdetectionprobabilityforpri-maryuserdetection..............................45Figure2.7:Example2:Theoptimalthresholdsformalicioususerdetectionfor=103.Here,P0=0:25............................46Figure2.8:Example3:Theoverallfalsealarmrateandtheoverallmissdetectionprobabilityformalicioususerdetection.Here,P0=0:25and=103.47Figure3.1:Theproposedcognitiveradionetworkarchitecture.............51Figure3.2:TheDVB-T2framestructure........................52Figure3.3:Generationoftheproposedreferencesignals...............55Figure3.4:Thefalsealarmrate(FAR)andmissdetectionprobability(MDP)versusSNRforprimaryuserdetection.......................61Figure3.5:Thefalsealarmrate(FAR)andmissdetectionprobability(MDP)versusSNRformalicioususerdetection......................62xiFigure4.1:Theproposedenergyharvestingscheme..................67Figure4.2:Theachievablesum-ratecomparisonoftheproposedOFDM-baseden-ergyharvestingschemeandthatofitsnon-energyharvestingcounterpartversustPUandMUtransmitpowerlevels.............83Figure4.3:TheoptimalandsuboptimalpowersplittingratiosversusPUtransmitpower.....................................84Figure5.1:Theproposedrelay-assistednetworkarchitecture.............89Figure5.2:Theproposed2-hoproutingtopology....................90Figure5.3:Example1:ThesystemblockingprobabilityPBversusrBS.Here,rBS;Opt=225:36m..............................103Figure5.4:Example2:ClassdistributionandclassblockingprobabilityversusMjusingrBS;Opt=225:36m..........................104xiiKEYTOABBREVIATIONSIoTInternetofThingsmmWaveMillimeterWaveCRCognitiveRadioDSADynamicSpectrumAccessPUPrimaryUserSUSecondaryUserSUCSecondaryUserCoordinatorMUMaliciousUserPUEAPrimaryUserEmulationAttackLFSRLinearFeedbackShiftRegisterAESAdvancedEncryptionStandardBPSKBinaryPhaseShiftKeyingRSSReceivedSignalStrengthLOSLineofSightDOADirectionofArrival8-VSBEightLevelVestigialSidebandDVB-T2SecondGenerationTerrestrialDTVStandardOFDMOrthogonalFrequencyDivisionMultiplexingOFDMAOrthogonalFrequencyDivisionMultipleAccessTDMATimeDivisionMultipleAccessCSIChannelStateInformationSNRSignaltoNoiseRatioSINRSignaltoInterferenceplusNoiseRatioQoSQualityofServiceBPBlockingProbabilityPPPPoissonPointProcessxiiiChapter1IntroductionInthischapter,weprovideabriefoverviewoftheIoTtechnologyandhighlightitsfun-damentalchallenges,whichmotivateustodevelopinnovativetechniquestoaddressthem.Thesetechniquesaresummarizedinthemaincontributionsofthisdissertation,whichin-cludefull-bandPUEAdetectionandmitigation,sub-bandPUEAdetectionandmitigation,energyharvestingandreliableinformationtransfer,andblockingprobabilityanalysisforrelay-assistedOFDMAnetworks.1.1Overview1.1.1MotivationsandProblemIdenTheInternetofThings(IoT),whichnetworksversatiledevicesforinformationexchange,remotesensing,monitoringandcontrol,isingpromisingapplicationsinnearlyeveryHowever,duetoitshighdensityandenormousspectrumrequirement,thepracticaldevelopmentofIoTtechnologyseemstobenotavailableuntilthereleaseofthelargemil-limeterwave(mmWave)band(30GHz-300GHz).Comparedtoexistinglowerbandsystems(suchas3G,4G),mmWavebandsignalsgenerallyrequirelineofsight(LOS)pathandfromseverefadingleadingtomuchsmallercoveragearea.Fornetworkdesignandmanagement,thisimpliesthat:(i)MmWavebandalonecouldnotsupporttheIoTnetworks,1buthastobeintegratedwiththeexistinglowerbandsystemsthroughsecureandespectrumsharing,especiallyinthelowerfrequencybands;and(ii)TheIoTnetworkswillhaveveryhighdensitynodedistribution,whichisatchallengeinnetworkdesign,especiallywiththescarceenergybudgetandthelowcostrequirementsofIoTapplications.Motivatedbytheseobservations,inthisdissertation,weconsiderthreeproblems:(1)Howtoachievesecureandespectrumsharing?(2)HowtoaccommodatetheenergylimitedIoTdevices?(3)HowtoevaluatetheQualityofService(QoS)inthehighdensityIoTnetworks?Weaimtodevelopinnovativetechniquesforthedesign,evaluationandmanagementoffutureIoTnetworksunderbothbenignandhostileenvironments.Thesetechniquesarehighlightedinthefollowingsubsection.1.1.2ResearchOverviewInthissubsection,weprovideanoverviewoftheproposedtechniquesthatcanbeappliedtothedesign,performanceevaluationandenhancementofIoTnetworks.Thesetechniquesinclude:DevelopingSecureandtSpectrumSharingSchemeintheHDTVBandsCognitiveradio(CR)isakeyenablingtechnologyforspectrumsharing,wheretheunoccupiedspectrumisidenforsecondaryusers(SUs),withoutinterferingwiththeprimaryuser(PU).AserioussecuritythreattotheCRnetworksisreferredtoasprimaryuseremulationattack(PUEA),inwhichamalicioususer(MU)emulatesthesignalcharacteristicsofthePU,therebycausingtheSUstoerroneouslyidentifytheattackerasthePU.Here,westartwithsingle-carriertransmission,wherethesignalisprocessedintheentirespectrumandhenceallowsfull-banddetection.Wecon-2siderfull-bandPUEAdetectionandproposeareliableAdvancedEncryptionStandard(AES)-assisteddigitalTV(DTV)scheme,whereanAES-encryptedreferencesignalisgeneratedattheDTVtransmitterandusedasthesyncbitsoftheDTVdataframes.Forprimaryuserdetection,weinvestigatethecross-correlationbetweenthereceivedsequenceandreferencesequence.Themalicioususerdetectioncanbeperformedbyinvestigatingtheauto-correlationofthereceivedsequence.Wefurtherdevelopasecureandtspectrumsharingschemeinmulti-carrierwirelessnetworks.Motivatedbytheprevalenceoftheorthogonalfrequencydivisionmultiplexing(OFDM)-basedsystems,weconsidersub-bandmalicioususerdetectionandproposeasecureAES-basedDTVscheme,wheretheexistingreferencesequenceusedtogeneratethepilotsymbolsintheDVB-T2framesisencryptedusingtheAESalgorithm,andtheresultedsequenceisexploitedforaccuratedetectionoftheautho-rizedPUandtheMU.Basedonthediscussionsabove,potentially,theproposedschemescanbeapplieddirectlytofutureIoTnetworksformorecientspectrumsharingwiththeexistinglowerbandsystems.DevelopingEnergytSchemesWeproposeanenergyttransmis-sionschemeinCRnetworksusingenergyharvesting,whereeachSUcanperforminformationreceptionandenergyharvestingsimultaneously.Weperformsum-rateop-timizationfortheSUsunderPUEA.Itisobservedthatthesum-rateoftheSUnetworkcanbeimprovedtlywiththeenergyharvestingtechnique.Inadditiontothejammingscenario,energyharvestingcanbeappliedinthebenignenvironmentaswelltoresolvetheenergyscarcityintheIoTnetworks.3DevelopingInnovativePerformanceAnalysisMethodologiesTakingthespa-tialrandomnessoftheIoTnetworkintoconsideration,weinvestigateblockingprobabil-ityinrelay-assistedorthogonalfrequencydivisionmultipleaccess(OFDMA)networksusingstochasticgeometry.Moresp,wemodeltheinter-cellinterferencefromtheneighboringcellsateachtypicalnode,andderivethecoverageprobabilityinthedownlinktransmissions.Basedontheirdataraterequirements,weclassifytheincom-ingusersintotclasses,andcalculatetheBPusingthemulti-dimensionallossmodel.WeshowthattheBPcanbereducedbyexploitingrelay-assistedtransmissions.1.2RevisitoftheExistingWork1.2.1PrimaryUserEmulationAttackAlongwiththeever-increasingdemandinhigh-speedwirelesscommunications,spectrumscarcityhasbecomeaseriouschallengetotheemergingwirelesstechnologies.Inlicensednetworks,theprimaryusersoperateintheirallocatedlicensedbands.Itisobservedthatthelicensedbandsaregenerallyunderutilizedandtheiroccupationtemporallyandgeographicallyintherangeof15%to85%[1].Cognitiveradionetworks[2{4]provideapromisingsolutiontothespectrumscarcityandunderutilizationproblems[5].CRnetworksarebasedondynamicspectrumaccess(DSA),wherethesecondaryusersareallowedtosharethespectrumwiththeprimaryusers.Thespectrumsharingismadeundertheconditionthatthesecondaryusersdonotinterferewiththeprimarysystem'sTheCRsidentifytheunusedbands(whitespaces)throughspectrumsensing[4],thenutilizetheidlebandsfordatatransmissions.Thespectrumsensingfunctioniscontinuouslyperformed.IfaprimarysignalisdetectedinthebandthataCRoperatesin,thentheCRmustevacuate4thatbandandoperateinanotherwhitespace[6].TheCRsystemisvulnerabletomaliciousattacksthatcoulddisruptitsoperation.Awell-knownmaliciousattackistheprimaryuseremulationattack[7{9].InPUEA,malicioususersmimictheprimarysignalovertheidlefrequencyband(s)suchthattheauthorizedsecondaryuserscannotusethecorrespondingwhitespace(s).Thisleadstolowspectrumutilizationandtcognitivenetworkoperation.PUEAhasattractedconsiderableresearchattentioninliterature.In[10],ananalyticalmodelfortheprobabilityofsuccessfulPUEAbasedontheenergydetectionwasproposed,wherethereceivedsignalpowerismodeledasalog-normallydistributedrandomvariable.Inthisapproach,alowerboundontheprobabilityofasuccessfulPUEAisobtainedus-ingMarkovinequality.In[11],anonparametricBayesianapproach,calledDECLOAK,wasinvestigatedtoidentifyPUEA.Theideaofthisapproachistousesomeofthetransmit-tedsignalparametersasattoidentifytheactualprimaryusers,andhencetheattackers.SeveralothermethodshavebeenproposedtodetectanddefendagainstPUEA.In[12],atransmittervnscheme(localization-baseddefense)wasproposedtodetectPUEA.In[13]and[14],theauthorsproposedareceivedsignalstrength(RSS)-baseddefensetechniquetodefendagainstPUEA,wheretheattackerscanbeidenbycomparingthereceivedsignalpoweroftheprimaryuserandthesuspectattacker.AWald'ssequentialprobabilityratiotest(WSPRT)waspresentedtodetectPUEAbasedonthereceivedsignalpowerin[15].AsimilarstrategywasusedtodetectPUEAinfadingwirelessenvironmentsin[16].In[17],acooperativesecondaryusermodelwasproposedforprimaryuserdetectioninthepresenceofPUEA.Inthisapproach,thedecisionwhethertheprimaryuserispresentorabsentisbasedontheenergydetectionmethod.In[18],theauthorsproposedcompressivesensing5Figure1.1:ApossiblescenariofortheattackerstoavoidPUEAdetectionapproachesbasedonthelocationand/ortheenergylevelofthereceivedsignal.Forexample,MU1canproducethesameDOAandcomparablereceivedpowerlevelastheprimaryuser,whileMU2canproducecomparablereceivedpowerlevelastheprimaryuser.basedapproachtodistinguishwhetherthesignalstransmittedarefromtheprimaryusersormalicioususers.InmobileCRnetworks,theauthorsin[19]proposedamethodtodetectPUEAonmobileprimaryusersbyexploitingthecorrelationsbetweenradiofrequency(RF)signalsandacousticinformationtoverifytheexistenceofmaliciouswirelessmicrophones.Recently,in[20],locationreliabilityandmaliciousintentionwereexploredtodetectprimaryuserandmalicioususerinmobileCRnetworks.Inmostoftheseexistingapproaches,thedetectionofPUEAismainlybasedonthepowerleveland/orthedirectionofarrival(DOA)ofthereceivedsignal.Thebasicideaisthat:giventhelocationsoftheprimaryTVstations,thesecondaryusercandistinguishtheactualprimarysignalfromthemalicioususer'ssignalbycomparingthepowerleveland/ortheDOAofthereceivedsignalwiththatoftheauthorizedprimaryuser'ssignal.Themajorlimitationwithsuchapproachesisthat:theywouldfailwhenamalicioususer6isatalocationwhereitproducesthesameDOAand/orcomparablereceivedpowerlevelasthatoftheactualprimarytransmitter,asshowninFigure1.1(seethepositionsofMU1andMU2).1.2.2EnergyHarvestingandReliableInformationTransferAlongwiththeemergingenergy-constrainedIoTnetwork,whichbasicallyreliesontransceiverswithlimitedpowercapabilities,wirelessenergytransferhasattractedmoreresearchinterestinrecentyears[21{35].Inliterature,therearethreemainwirelesspowertransfertechniques,namely,magneticinductivecoupling,magneticresonancecoupling,andRFenergyharvesting.Inthisdissertation,wefocusontheRFenergyharvestingtechniques.Magneticinductivecouplingisbasedonthemagneticinductionprinciple,whereavaryingcurrentattheprimary/transmittercoilgeneratesavaryingmagneticwhichthencanbecapturedbyasecondary/receivercoilandconvertedthemagneticbacktoelectricity.Thepowertransferdependsonphysicalseparationbetweenthetwocoilsandtheirqualityfactor.In[21],itwasshownthat,fordistancesrangefrom2cmto8cm,theciencyofthemagneticinductivecouplingvariesroughlyfrom65%to5%,respectively.Magneticresonancecouplingissimilartotheinductivecoupling,butwithanadditionofacapacitanceonboththeprimaryandsecondarycoils,whichtheyarethenreferredtoasresonators.Toachievehighpowertransfer,thetworesonatorsshouldworkonthesamefrequency[22].Theauthorsin[23]wereabletoacheivepowertransferfrom30%toabove90%withdistancesvaryfrom2:25mto0:75m,respectively,usingmagneticresonancecouplingmethod.EnergyharvestingfromambientRFsignalshasgainedconsiderableattentioninbothindustrialandacademicTheconceptofsimultaneouswirelessinformationtransfer7andenergyharvestingwasproposedin[24],wherethefundamentalbetweeninformationreceivingandpowertransferwasstudiedfromaninformationtheoryperspective.Inliterature[25,26],therehavebeentworepresentativereceiverarchitecturesforenergyharvesting:power-splittingandtime-switching.Inthepower-splittingbasedapproach,apowersplitterisemployedattheRFbandatthereceiver.ThereceivedRFsignalissplitintotwoparts:oneforenergyharvestingandtheotherforinformationprocessing.Inthetime-switchingapproach,thereceivedsignalisdividedintotwopartsinthetimedomain.Thepowertransferofenergyharvestingtechniquesdependsmainlyonthereceivedsignalstrength.In[27],itwasshownthatanof0:4%,18:2%andabove50%canbeachievedwithRFenergytransferwithinputpower40dBm,5dBm,and20dBm,respectively.Moredetailsonperformanceofwirelessenergytransfertechniquescanbefoundin[28].Recently,energyharvestinghasbeenwidelydiscussedinOFDM-basedsystems[29],wirelessrelaynetworks[30],andcognitiveradionetworks[31{35].Forexample,theauthorsin[31]investigatedtheoptimaltransmissionandenergyharvestingpolicytomaximizethesecondarynetworkthroughputinacognitiveradionetwork,wherethesecondaryusersop-portunisticallyaccessthespectrumorharvestenergyfromtheambientRFsignalsfromprimaryusers.In[32],theauthorsstudiedthechannelaccessprobleminCRnetworksinwhichasecondaryusercanchoosetoaccessthechannelfordatatransmissionorenergyharvesting.Toobtainthechannelaccess,theypresentedanoptimizationformulationbasedonMarkovdecisionprocess.Generally,PUEAwasnotconsideredintheseworks[31{35].However,PUEAises-sentiallyjamminginterferencefortheSUs,andpotentiallytheperformanceofSUscanbeimprovedtlybyexploitingPUEAasanextraenergyresource.Inadditiontothe8jammingscenario,energyharvestingcanbeappliedinthebenignenvironmentaswelltoresolvetheenergyscarcityintheIoTnetworks.1.2.3QoSPerformanceMeasureinIoTNetworksIndesigningandanalyzingwirelessnetworks,blockingprobabilityhasbeenusedasaveryimportantmetricinevaluatingtheQoSofthenetwork.Blockingprobabilityistheprob-abilitythatanarrivinguserisdeniedofserviceduetotnetworkresources.Inliterature,blockingprobabilityhasbeeninvestigatedfortheOFDMAnetworksfromentperspectives.Toprovidemorereliableservicesforuserswithlowreceivedsignalstrength,relaystationsareoftendeployedalongwiththebasestationstoextendthecoverageareaandimprovetheQoS.Asaresult,blockingprobabilityhasbeenstudiedinbothsingle-hopandmulti-hopnetworks.Somerepresentativeworkonblockingprobabilityforthesingle-hopOFDMAnetworkscanbefoundin[36{38].In[36],theauthorstriedtomodelthesubcarrier-allocationsystemusingthebatch-arrivalmodelMX=M=c=c.Thismodelisnottlyaccurateduetotheassumptionthatthesubcarriersareallocatedinbulksbutreleasedonebyone.Therefore,in[37],amorerealisticmodelknownasmulti-classErlanglossmodelwasproposed.Motivatedbythefactthatpowershouldbeconsideredinadditiontosubcarriersasasystemresource,amoregeneralmodelwasproposedin[38],whereboththepowerandthesubcarriersareregardedasthesystemresources.Blockingprobabilityformulti-hopOFDMAnetworksisattractingmoreresearchatten-tionalongwiththeemerginghighdensityIoTnetworks.Forexample,in[39],theauthorsproposedasystemmodeltoevaluatetheblockingprobabilityfortherelay-basedOFDMAcellularnetworks.Intheirmodel,eachcellconsistsofabasestation,locatedatthecenterof9thecell,andissurroundingbysixrelaystationswithdeterministic(known)locations.How-ever,aspointedoutin[40],resultsbasedonsuchhighlyidealizedmodelsgenerallymaynotbeveryaccurate.Ontheotherhand,itwasalsoshownin[40]thatmodelingthelocationsofthebase/relaystationsstochasticallyaccordingtoaPoissonpointprocess(PPP)depictstherealityandcanachievemorereliableandaccurateresultscomparedtoitsidealizedcoun-terparts.Moreclosely,stochasticgeometryhasbeenwidelyappliedinmodelingwirelessnetworksinrecentyears[41{44].Forinstance,in[41],stochasticgeometrywasusedtoan-alyzerelay-aidedtwo-hopnetworks.In[42],itwasusedtoanalyzemulti-hoptransmissioninad-hocnetworks.However,noworkhasbeendoneinliteraturetoinvestigateblockingprobabilityinrelay-assistedOFDMAnetworksusingstochasticgeometry.Moreover,averychallengingproblemincalculatingtheblockingprobabilityistoobtainthedistributionofthegroupsizeofthesubcarriersthatanarrivinguserneedstoitsdatarate.Inmostoftheexistingwork[36{38],thisdistributionisassumedtobeknown,usuallyassumedtobeuniform.Thismaynotbeaccurate,especiallyforlargegroupsizes,whichactuallyhaveverylowrequestprobability.1.3SummaryofDissertationContributionsThemaincontributionsofthisdissertationaresummarizedinthefollowing.InChapter2,wedevelopasecureandspectrumsharingschemeinsingle-carrierwirelessnetworks.Weconsiderfull-bandPUEAdetectioninsingle-carrierATSC1.0DTVsystem,wheretheprimaryuser'ssignalisprocessedintheentirespectrum.WeproposeareliableAES-assistedDTVscheme,whereanAES-encryptedreferencesignalisgeneratedattheTVtransmitterandusedasthesyncbitsoftheDTVdataframes.Byallowingashared10secretbetweenthetransmitterandthereceiver,thereferencesignalcanberegeneratedatthereceiverandusedtoachievefull-bandidenofauthorizedprimaryusers.Moreover,whencombinedwiththeanalysisontheauto-correlationofthereceivedsignal,thepresenceofthemalicioususercanbedetectedaccuratelynomattertheprimaryuserispresentornot.Theproposedschemecombatsprimaryuseremulationattacks,andenablesmorerobustsystemoperationandtspectrumsharing.Thectivenessoftheproposedapproachisdemonstratedthroughboththeoreticalanalysisandsimulationexamples.ItisshownthatwiththeAES-assistedDTVscheme,theprimaryuser,aswellasmalicioususer,canbedetectedwithhighaccuracyandlowfalsealarmrateunderprimaryuseremulationattacks.InChapter3,wefurtherdevelopasecureandspectrumsharingschemeinmulti-carrierwirelessnetworks.Unlikesingle-carriertransmissionschemes,whichallowonlyfull-bandPUEAdetection,multi-carrierschemesprovidepotentialyinperformingsub-bandPUEAdetectioninCRnetworks.MotivatedbythisfactandtheprevalenceoftheOFDM-basedDTVstandard,weconsidersub-bandmalicioususerdetectioninOFDM-basedCRnetworkunderPUEA.Here,theCRnetworkconsistsofprimaryusers,secondaryusercoordinators(SUCs),secondaryusers,andmalicioususers.TheSUCsaredesignedtoperformspectrumsensingandsub-bandPUEAdetection,andthencoordinatetheSUsforcollision-freetransmissionoverthewhitespaces.WeproposearobustandtAES-basedDTVscheme,wheretheexistingreferencesequenceusedtogeneratethepilotsymbolsintheDVB-T2framesisencryptedusingtheAESalgorithm,andtheresultedsequenceisexploitedforaccuratedetectionoftheauthorizedprimaryuserandthemalicioususer.AlongwiththeOFDM-basedtransceiverstructureandpilotsymbolallocationschemeintheDVB-T2standard,wecandetectthepresenceofPUEAovereach3-subcarriersub-band.However,theproposedapproachcanbeusedtodetectPUEAovereachsinglesubcarrierif11thepreamblesymbolsintheDVB-T2standardcanbeadjustedtocovereverysubcarrier.TheperformanceoftheproposedPUEAdetectionapproachisevaluatedthroughfalsealarmrateandmissdetectionprobability.OuranalysisindicatesthattheproposedapproachcandetectPUEAwithhighaccuracy.Potentially,theproposedschemecanbeapplieddirectlytofutureIoTnetworksformoretspectrumsharingwiththeexistinglowerbandsystems.InChapter4,wedevelopanenergytransmissionschemeinCRnetworksusingenergyharvesting.Here,themaincontributionistwofold.First,wecomeupwithantivecommunicationschemeforthesecondaryusersunderPUEAbyexploitingtheenergyharvestingtechniques.UsingPUEAasanextrapowerresource,wepresentatransmit-tingschemefortheSUssuchthateachSUcanperforminformationreceptionandenergyharvestingsimultaneously.Weperformsum-rate(downlinktransmissionrateplusuplinktransmissionrate)optimizationfortheSUsunderPUEA.Astheoptimalsolutionisbasedonmulti-dimensionalexhaustivesearch,wefurtherproposeasuboptimalschemewithclosed-formsolution.Oursimulationsdemonstratethattheperformanceofthesuboptimalschemeisveryclosetothatoftheoptimalsolution.Moreover,asexpected,sigtperformanceimprovementcanbeobservedwhencomparingwithnon-energyharvestingsystems.Inad-ditiontothejammingscenario,energyharvestingcanbeappliedinthebenignenvironmentaswelltoresolvetheenergyscarcityintheIoTnetworks.Second,weevaluatetheworst-casePUEAinterferenceintermsofminimizingthesum-rateforthesecondaryusers.Weprovethatforthesecondaryusers,equalpowerinterferencefromthemalicioususeristheworstinterferenceforweakjamming(i.e.,whenthereceivedsignalpowerismuchlargerthanthereceivedjammingpower),andnearlytheworstin-terferenceforstrongjamming(i.e.,whenthereceivedjammingpowerismuchlargerthan12thereceivedsignalpowerandthenoisepower),whilethechannelstateinformation(CSI)assistedinterferenceistheworst-caseinterferenceforstrongjamminginthehighsignal-to-noiseratio(SNR)region.Moreover,weshowthattheresultedsum-rateperformancegapbetweenequalpowerinterferenceandCSI-assistedinterferenceissmall.Theseresultsindi-catethatforpracticalsystems,duetotheabsenceofCSIinformation,theworstjammingfortheSUsiswhenthemalicioususerperformsequalpowerallocationoverallthewhitespacesubcarriers.InChapter5,takingthespatialrandomnessoftheIoTnetworkintoconsideration,weinvestigatecoverageprobabilityandblockingprobabilityinrelay-assistedOFDMAnetworksusingstochasticgeometry.InOFDMAnetworks,usersareassigneddtnumberofresourceelements(subcarriers)tomeettheirraterequirements.Therefore,theincominguserscanbeintotclassesbasedontheirsubcarrierrequirements.Averychallengingproblemincalculatingtheblockingprobabilityistoobtainthedistributionofthegroupsizeofthesubcarriersthatanarrivinguserneedstoitsdatarate.Inmostoftheexistingwork[36{38],thisdistributionisassumedtobeknown,usuallyassumedtobeuniform.Thismaynotbeaccurate,especiallyforlargegroupsizes,whichactuallyhaveverylowrequestprobability.Here,wederivethedistributionofuserrequiredsubcarriergroupsizeandthenobtaintheuserblockingprobabilitybasedonthat.Moresp,wemodeltheinter-cellinterferencefromtheneighboringcellsatatypicalnode.Second,wederivethecoverageprobability(i.e.,theprobabilityofsuccessfultransmission)inthedownlinktransmissions,includingboththedirectandrelay-basedtransmissions.Third,weclassifytheincomingusersintotclassesbasedontheirsubcarrierrequirements.Thesystemundercon-siderationcanbemodeledusingmulti-dimensionalMarkovchains.Finally,wecalculatethe13blockingprobabilityusingthemulti-dimensionallossmodel.Itisshownthattheblockingprobabilitycanbereducedbyexploitingrelay-assistedtransmissions.InChapter6,wesummarizetheconclusionsandprovidesomeoutlinesforfuturework.14Chapter2Full-bandPUEADetectionandMitigationInthischapter,weconsiderfull-bandPUEAdetectionincognitiveradionetworksoperatinginthewhitespacesoftheDTVband.WeproposeareliableAES-assistedDTVscheme,whereanAES-encryptedreferencesignalisgeneratedattheTVtransmitterandusedasthesyncbitsoftheDTVdataframes.Byallowingasharedsecretbetweenthetransmitterandthereceiver,thereferencesignalcanberegeneratedatthereceiverandusedtoachievefull-bandidencationofauthorizedprimaryusers.Moreover,whencombinedwiththeanalysisontheauto-correlationofthereceivedsignal,thepresenceofthemalicioususercanbedetectedaccuratelynomattertheprimaryuserispresentornot.ItshouldbenotedthattheAES-encryptedreferencesignalisalsousedforsynchronizationpurposesattheauthorizedreceivers,inthesamewayastheconventionalsynchronizationsequence.Weanalyzetheenessoftheproposedapproachthroughboththeoreticalanalysisandsimulationexamples.ItisshownthatwiththeAES-assistedDTVscheme,theprimaryuser,aswellasmalicioususer,canbedetectedwithhighaccuracyunderprimaryuseremulationattacks.Itshouldbeemphasizedthattheproposedschemerequiresnochangesinhardwareorsystemstructureexceptofaplug-inAESchip.Potentially,theproposedschemecanbeapplieddirectlytofutureIoTnetworksformoretspectrumsharingwiththeexisting15lowerbandsystems.2.1IntroductionCognitiveradionetworks[2{4]havereceivedconsiderableresearchattentionrecentlybecauseoftheirabilitytoalleviatethespectrumscarcityproblemduetotherapidgrowthinthewirelesscommunicationdevices.ThebasicideaoftheCRnetworksistoallowtheunlicenseduserstosharethefrequencyspectrumwiththelicensedusersundertheconditionthattheymustnotcauseharmfulinterferencetotheprimaryusers.Thespectrumsharingisperformedthroughspectrumsensing,wheretheCRssensethespectrumtoidentifytheunusedbands(whitespaces)fordatatransmission.IfaprimarysignalisdetectedinthebandthataCRoperatesin,thentheCRmustevacuatethatbandandoperateinanotherwhitespace.AserioussecuritythreattotheCRnetworksisknownasprimaryuseremulationattack[7{9],wherethemalicioususersemulatetheprimarysignalovertheidlefrequencyband(s)suchthatthesecondaryuserscannotusethecorrespondingwhitespace(s).SeveralapproacheshavebeenproposedtodetectanddefendagainstPUEA[12{15].In[12],alocalization-basedtransmittervschemewasproposedtodetectPUEA.In[13]and[14],theauthorsproposedaRSS-basedtechniquetodefendagainstPUEA,wheretheattackerscanbeidenbycomparingthereceivedsignalpoweroftheprimaryuserandthesuspectattacker.AWald'ssequentialprobabilityratiotestwaspresentedtodetectPUEAbasedonthereceivedsignalpowerin[15].Intheseexistingapproaches,thedetectionofPUEAismainlybasedonthepowerleveland/orDOAofthereceivedsignal.Thebasicideaisthat:giventhelocationsoftheprimaryTVstations,thesecondaryusercandistinguishtheactualprimarysignalfromthemalicioususer'ssignalbycomparingthe16powerlevelandDOAofthereceivedsignalwiththatoftheauthorizedprimaryuser'ssignal.Themajorlimitationwithsuchapproachesisthat:theywouldfailwhenamalicioususerisatalocationwhereitproducesthesameDOAandcomparablereceivedpowerlevelasthatoftheactualprimarytransmitter.Inthischapter,weproposeareliableAES-assistedDTVscheme,whereanAES-encryptedreferencesignalisgeneratedattheTVtransmitterandusedasthesyncbitsoftheDTVdataframes.Byallowingasharedsecretbetweenthetransmitterandthereceiver,thereferencesignalcanberegeneratedatthereceiverandusedtoachieveaccurateidenofautho-rizedprimaryusers.Moreover,whencombinedwiththeanalysisontheauto-correlationofthereceivedsignal,thepresenceofthemalicioususercanbedetectedaccuratelynomattertheprimaryuserispresentornot.TheproposedapproachcanelycombatPUEAwithnochangeinhardwareorsystemstructureexceptofaplug-inAESchip,whichhasbeencommercializedandwidelyavailable[45{47].ItshouldbenotedthattheAES-encryptedreferencesignalisalsousedforsynchronizationpurposesattheauthorizedreceivers,inthesamewayastheconventionalsynchronizationsequence.Theproposedschemecombatsprimaryuseremulationattacks,andenablesmorerobustsystemoperationandtspectrumsharing.Thectivenessoftheproposedapproachisdemonstratedthroughboththeoreticalanalysisandsimulationexamples.ItisshownthatwiththeAES-assistedDTVscheme,theprimaryuser,aswellasmalicioususer,canbedetectedwithhighaccuracyandlowfalsealarmrateunderprimaryuseremulationattacks.172.2TheProposedAES-assistedDTVApproachInthissection,wepresenttheproposedAES-assistedDTVschemeforrobustandreliableprimaryandsecondarysystemoperations.WerstintroducethecurrentterrestrialdigitalTVsystem.Then,wediscussthetransmitterandthereceiverdesignsoftheproposedAES-assistedDTVscheme.Furthermore,weanalyzethedetectionproblemoftheproposedapproachusingcorrelation-basedmethods.Finally,wediscusssomepossibleconcernswiththeproposedAES-assistedDTVscheme,andprovidesomepracticalsolutions.2.2.1ABriefReviewoftheTerrestrialDTVSystemDTVistheadvancedtechnologyforenhancingthequalityandperformanceoftheanalogtelevisionbroadcasting.SeveralgreatbcanbegainedbytheadoptionoftheDTVsystemssuchasbetterpictureandsoundquality,lesstransmissionpower,andspectrum,whereuptosixchannelscanbroadcastsimultaneouslyoverthesamefrequencybandthatisusedbyoneanalogchannel[48].ManycountrieshaveswitchedfromtheanalogTVbroadcastingtothedigitalTVbyadoptingoneofthefourwidelyusedDTVbroadcastingstandards:AdvancedTelevisionSystemCommittee(ATSC),DigitalVideoBroadcasting-Terrestrial(DVB-T),TerrestrialIntegratedServicesDigitalBroadcasting(ISDB-T),andDigitalTerrestrialMultimediaBroadcasting(DTMB).IntheUnitedStates,theFederalCommunicationsCommission(FCC)hasadoptedtheATSCstandardastheDTVterrestrialbroadcasts.In1996,theU.S.governmentallowedtheTVcompaniestobroadcastdigitalsignalsalongwiththeanalogbroadcasting.By2009,theFCChasannouncedthatdigitalTVbroadcastingismandatoryintheU.S.IntheATSCstandard,eight-levelvestigialsideband(8-VSB)modulationisusedfor18Figure2.1:8-VSBsignalframestructure.transmittingdigitalsignalsaftertheyarepartitionedintoframes[49].Theframestructureofthe8-VSBsignalisillustratedinFigure2.1.Eachframehastwodataandeachdatadhas313datasegments.Thedatasegmentofeachdataisusedforframesynchronizationandchannelestimationatthereceiver[49],[50].Theremaining624segmentsareusedfordatatransmission.Eachdatasegmentcontains832symbols,including4symbolsusedforsegmentsynchronization.Thesegmentsynchronizationbitsareidenticalforalldatasegments.Eachsegmentlasts77:3,hencetheoveralltimedurationforoneframe,whichhas626segments,is62677:3=48:4ms[49].2.2.2AES-assistedDTVTransmitterTheDTVtransmitterobtainsthereferencesignalthroughtwosteps:generatingapseudo-randombinarysequence(PRBS),thenencryptingthesequencewiththeAESalgo-19Figure2.2:Generationofthereferencesignal.rithm.Moresp,apseudo-randombinarysequenceisgeneratedusingaLinearFeedbackShiftRegister(LFSR)1withasecureinitializationvector(IV).Maximum-lengthLFSRsequencescanbeachievedbytappingtheLFSRsaccordingtoprimitivepolynomials.Themaximumsequencelengththatcanbeachievedwithaprimitivepolynomialofdegreemis2m1.Withoutlossofgenerality,amaximum-lengthsequenceisassumedthroughoutthischapter.Oncethemaximum-lengthsequenceisgenerated,itisusedasaninputtotheAESencryptionalgorithm,asillustratedinFigure2.2.Weproposethata256-bitsecretkeybeusedfortheAESencryptionsothatthemaximumpossiblesecurityisachieved.SecurityanalysiswillbeprovidedinSection2.4.Denotethepseudo-randombinarysequencebyx,thentheoutputoftheAESalgorithmisusedasthereferencesignal,whichcanbeexpressedas:s=E(k;x);(2.1)wherekistheencryption/decryptionkey,andE(;)denotestheAESencryptionoperation.ThetransmitterthenplacesthereferencesignalsinthesyncbitsoftheDTVdatasegments.ThesecretkeycanbegeneratedanddistributedtotheDTVtransmitterandreceiver1Anyotherpseudo-randomgeneratorscanbeusedaswell.20fromatrustedthirdpartyinadditiontotheDTVandtheCRuser.ThethirdpartyservesastheauthenticationcenterforboththeprimaryuserandtheCRuser,andcancarryoutkeydistribution.Topreventimpersonationattack,thekeyshouldbetimevarying[51].2.2.3AES-assistedDTVReceiverThereceiverregeneratestheencryptedreferencesignal,withthesecretkeyandIVthataresharedbetweenthetransmitterandthereceiver.Acorrelationdetectorisemployed,whereforprimaryuserdetection,thereceiverevaluatesthecross-correlationbetweenthereceivedsignalrandtheregeneratedreferencesignals;formalicioususerdetection,thereceiverfurtherevaluatestheauto-correlationofthereceivedsignalr.Thecross-correlationoftworandomvariablesxandyisas:Rxy=<x;y>=Efxyg:(2.2)UnderPUEA,thereceivedsignalcanbemodeledas:r=s+m+n;(2.3)wheresisthereferencesignal,misthemalicioussignal,nisthenoise,andarebinaryindicatorsforthepresenceoftheprimaryuserandmalicioususer,respectively.Moresp,=0or1meanstheprimaryuserisabsentorpresent,respectively;and=0or1meansthemalicioususerisabsentorpresent,respectively.212.2.3.1DetectionofthePrimaryUserTodetectthepresenceoftheprimaryuser,thereceiverevaluatesthecross-correlationbe-tweenthereceivedsignalrandthereferencesignals,i.e.,Rrs=<r;s>=<s;s>+<m;s>+<n;s>=˙2s;(2.4)where˙2sistheprimaryuser'ssignalpower,ands,m,nareassumedtobeindependentwitheachotherandareofzeromean.Dependingonthevalueofin(2.4),thereceiverdecideswhethertheprimaryuserispresentorabsent.Assumingthatthesignalsareergodic,thentheensembleaveragecanbeapproximatedbythetimeaverage.Here,weusethetimeaveragetoestimatethecross-correlation.Theestimatedcross-correlation^Rrsisgivenby:^Rrs,NXi=1risiN;(2.5)whereNisthereferencesignal'slength,siandridenotetheithsymbolofthereferenceandreceivedsignal,respectively.Todetectthepresenceoftheprimaryuser,thereceivercomparesthecross-correlationbetweenthereferencesignalandthereceivedsignaltoathreshold.Wehavetwocases:1.Ifthecross-correlationisgreaterthanorequalto,thatis:^Rrs(2.6)22thenthereceiverconcludesthattheprimaryuserispresent,i.e.,=1.2.Ifthecross-correlationislessthan,thatis:^Rrs<(2.7)thenthereceiverconcludesthattheprimaryuserisabsent,i.e.,=0.Thisdetectionproblemcanbemodeledasabinaryhypothesistestproblemwiththefollowingtwohypotheses:H0:theprimaryuserisabsent(^Rrs<)H1:theprimaryuserispresent(^Rrs)Ascanbeseenfrom(2.4),thecross-correlationbetweenthereferencesignalandthereceivedsignalisequalto0or˙2s,incasewhentheprimaryuserisabsentorpresent,respectively.Followingtheminimumdistancerule,wechoose=˙2s=2asthethresholdforprimaryuserdetection.2.2.3.2DetectionoftheMaliciousUserFormalicioususerdetection,thereceiverfurtherevaluatestheauto-correlationofthere-ceivedsignalr,i.e.,Rrr=<r;r>=2<s;s>+2<m;m>+<n;n>=2˙2s+2˙2m+˙2n;(2.8)where˙2mand˙2ndenotethemalicioususer'ssignalpowerandthenoisepower,respectively.Basedonthevalueof,canbedeterminedaccordinglythrough(2.8).Wehavethe23followingcases:Rrr=8>>>>>>>>><>>>>>>>>>:˙2s+˙2m+˙2n;=1;=1˙2s+˙2n;=1;=0˙2m+˙2n;=0;=1˙2n;=0;=0(2.9)Assumingergodicsignals,wecanusethetimeaveragetoestimatetheauto-correlationasfollows:^Rrr,NXi=1ririN:(2.10)Here,wecanmodelthedetectionproblemusingfourhypotheses,denotedbyH,where;2f0;1g:H00:themalicioususerisabsentgiventhat=0H01:themalicioususerispresentgiventhat=0H10:themalicioususerisabsentgiventhat=1H11:themalicioususerispresentgiventhat=1Inpracticalscenarios,however,weonlyhaveanestimatedvalueof,denotedas^.Weestimateafterweobtain^.Todothis,thereceivercomparestheauto-correlationofthereceivedsignaltotwodthresholds0and1basedonthepreviouslydetected^.Moresp,thereceivercomparestheauto-correlationofthereceivedsignalto024when^=0,andto1when^=1.Thatis:8>>>>>>>>><>>>>>>>>>:^H00:^Rrr<0;giventhat^=0;(^=0)^H01:^Rrr0;giventhat^=0;(^=1)^H10:^Rrr<1;giventhat^=1;(^=0)^H11:^Rrr1;giventhat^=1;(^=1)(2.11)Theperformanceofthedetectionprocessfortheprimaryuserandmalicioususeriseval-uatedthroughthefalsealarmratesandthemissdetectionprobabilities,aswillbediscussedinSection2.3.2.2.4FurtherDiscussionsThenatureoftheCRnetworksoperation,whichisbasedonthecoexistenceofprimaryusersandsecondaryusers,makesitvulnerabletohostileattackssuchasPUEA.SeveralapproacheshavebeenproposedtodetectPUEA,whichcanbecategorizedintotwoclasses:(i)energylevelandDOAbasedapproaches[12{17],and(ii)userauthenticationapproaches[52,53].InChapter1,werevisitedsomeenergylevelbasedapproaches,anddiscussedtheirmajorlimitations.Thatis,theywouldfailwhenamalicioususerisatalocationwhereitproducesthesameDOAand/orcomparablereceivedpowerlevelasthatoftheactualprimarytransmitter,asshowninFigure1.1.Theprimaryuserandsecondaryuserdetectionapproachesproposedinthischaptercanelyovercomethisdrawback.Someotheruserauthenticationbasedtechniqueshavealsobeenproposedsuchasin[52,53].In[52],apublickeycryptographymechanismisusedbetweenprimaryusersand25secondaryusers,suchthatthesecondaryuserscanidentifytheprimaryusersaccuratelybasedontheirpublickeys.Apossibleconcernwiththisschemeisthatpublickeybasedapproachesgenerallyhavehighcomputationalcomplexity.In[53],atwo-stageprimaryuserauthenticationmethodwasproposed:(i)generatetheauthenticationtagfortheprimaryuserusingaone-wayhashchain,and(ii)embedthetagintheprimaryuser'ssignalthroughconstellationshift.Sincetheauthenticationtagissuperimposedovertheprimaryusertrans-mittedsymbols,itintroducessomedistortionstotheprimaryusersignals,andissensitivetonoise.Comparingwiththeexistinguserauthenticationbasedapproaches,ourapproachismoretandhashigherdetectionaccuracy.Althoughuserauthenticationapproachesaregenerallymorereliableundervariousattackscenariosandgenerallyhavenoassumptionsontheprimaryuser'stransmissionpowerorlocation,theycanonlybeappliedtodetectthepresenceoftheprimaryuserandthemalicioususerbutnotthewhitespacesinthespectrum.Amoreeandpracticalsolutionforthisproblemwouldbetocombinetheproposedapproachwiththeenergyleveldetectionapproaches.Inthiscase,boththeprimaryuserandmalicioususer,aswellasthewhitespaces,canbeaccuratelyidenTocompletelyresolvethisproblem,theprimaryuserneedstousemulti-carriersystemsuchasOFDM,whereeachsubcarrieroperatesinaparticularsub-bandintheallocatedfrequencyspectrum.Withthis,itispossibletodetecttheprimaryuserandmalicioususerineachsub-bandusingtheproposedscheme,whichwewillconsiderinChapter3.262.3AnalyticalEvaluationoftheProposedAES-assistedDTVApproachInthissection,weanalyzethedetectionperformanceofthetheproposedAES-assistedDTVapproach.First,weevaluatethesystemperformanceforprimaryuserdetection.Then,weanalyzetheenessoftheproposedAES-assistedDTVschemeindetectingmaliciousnodes.2.3.1AnalyticalEvaluationofPrimaryUserDetectionWeassumethatthedetectionoftheprimaryuserhasafalsealarmratePfandamissdetectionprobabilityPm,respectively.ThefalsealarmratePfistheconditionalprobabilitythattheprimaryuserisconsideredtobepresent,whenitisactuallyabsent,i.e.,Pf=Pr(H1jH0):(2.12)ThemissdetectionprobabilityPmistheconditionalprobabilitythattheprimaryisconsid-eredtobeabsent,whenitispresent,i.e.,Pm=Pr(H0jH1):(2.13)Ascanbeseenfrom(2.5),^RrsistheaveragedsummationofNrandomvariables.SinceNislarge,thenbasedonthecentrallimittheorem,^RrscanbemodeledasaGaussianrandomvariable.Moresp,underH0,^Rrs˘N(0;˙20),andunderH1,^Rrs˘N(1;˙21),where0,˙0,and1,˙1canbederivedasfollows.27UnderH0,thereceivedsignalisrepresentedasri=mi+ni,wheremiistheithmalicioussymbol,andni˘N(0;˙2n).Then,themean0canbeobtainedas:0=1NE8<:NXi=1(mi+ni)si9=;=1NE8<:NXi=1(misi+nisi)9=;=1N"NXi=1(EfmigEfsig+EfnigEfsig)#=0:(2.14)Thevariance˙20canbeobtainedas:˙20=Enj^Rrsj2oj0j2=1N2E8<:NXi=1(mi+ni)siNXj=1(mj+nj)sj9=;=1N2E8<:NXi=1NXj=1(2mimjsjsi+ninjsjsi)9=;=1N2"NXi=1(2Efjmij2gEfjsij2g+Efjnij2gEfjsij2g)#=1N2˙2s˙2m+˙2s˙2n):(2.15)Similarly,underH1,thereceivedsignalisrepresentedasri=si+mi+ni,andthe28mean1canbeobtainedasfollows:1=1NE8<:NXi=1(si+mi+ni)si9=;=1NE8<:NXi=1(sisi+misi+nisi)9=;=1N"NXi=1(Efjsij2g+EfmigEfsig+EfnigEfsig)#=˙2s;(2.16)and˙21canbeobtainedas:˙21=Enj^Rrsj2oj1j2=1N2E8<:NXi=1(si+mi+ni)siNXj=1(sj+mj+nj)sj9=;(˙2s)2=1N2E8<:NXi=1NXj=1(sisisjsj+2mimjsjsi+ninjsjsi)9=;(˙2s)2=1N2"NXi=1(Efjsij4g+2Efjmij2gEfjsij2g+Efjnij2gEfjsij2g)+NXi=1NXj=1j6=iEfjsij2gEfjsjj2g#(˙2s)2=1N2N(Efj~sj4g+2˙2m˙2s+˙2n˙2s)+N(N1)(˙2s)2(˙2s)2=1NEfj~sj4g+2˙2m˙2s+˙2n˙2s(˙2s)2;(2.17)whereweassumethatEfjsij4g=Efj~sj4g8i.29Following(2.12),thefalsealarmratePfcanbeobtainedas:Pf=Prf^RrsjH0g=1p2ˇ˙01Ze(x0)22˙20dx=Q(0˙0):(2.18)Similarly,following(2.13),themissdetectionprobabilityPmcanbeobtainedas:Pm=Prf^Rrs<jH1g=1p2ˇ˙1Ze(x1)22˙21dx=1Q(1˙1):(2.19)Remark2.1AswillbeshownlaterinSection2.5,when=˙2s=2,bothPfandPmareessentiallyzero,andindependentoftheSNRvalues.TheunderlyingargumentisthatthedetectionoftheprimaryuserisbasedonRrs=˙2s(see(2.4)),whichisindependentofboth˙2mand˙2n.302.3.2AnalyticalEvaluationofMaliciousUserDetection2.3.2.1FalseAlarmRateandMissDetectionProbabilityforMaliciousUserDetection~Pf;0and~Pf;1asthefalsealarmratewhen^=0or^=1,respectively,~Pf;0=Pr(^H01j^H00);(2.20)~Pf;1=Pr(^H11j^H10):(2.21)Theoverallfalsealarmrateisgivenby:~Pf=^P0~Pf;0+(1^P0)~Pf;1;(2.22)where^P0istheprobabilitythat^=0,i.e.,^P0=(1Pf)P(=0)+PmP(=1):(2.23)AswillbeshowninSection2.4,withtheavalancheoftheAESalgorithm,thecross-correlationbetweenthereferencesignalandthereceivedsignalisalwaysaround˙2sor0,dependingonwhethertheprimaryuserispresentorabsent,respectively.Thatis,PfandPmarenegligible,aswillbedemonstratedlaterinSection2.5.Therefore,inthefollowing,weassumethat^=,andwedonotdistinguishbetween^H^andH;itfollowsthat^P0=P0=P(=0).Hence,theoverallfalsealarmrateisgivenby:~Pf=P0~Pf;0+(1P0)~Pf;1:(2.24)31Similarly,themissdetectionprobabilitiescanbeas~Pm;0and~Pm;1,whentheprimaryuserisabsentandpresent,respectively,i.e.,~Pm;0=Pr(H00jH01):(2.25)~Pm;1=Pr(H10jH11):(2.26)Theoverallmaliciousnodemissdetectionprobabilityisas:~Pm=P0~Pm;0+(1P0)~Pm;1:(2.27)Since^Rrristheaveragedsummationofalargenumberofrandomvariables,thenbasedonthecentrallimittheorem,^RrrcanbemodeledasaGaussianrandomvariable.Hence,wehave:8>>>>>>>>><>>>>>>>>>:^Rrr˘N(00;˙200);H00^Rrr˘N(01;˙201);H01^Rrr˘N(10;˙210);H10^Rrr˘N(11;˙211);H11(2.28)where00,˙00,01,˙01,10,˙10,and11,˙11canbederivedasfollows.UnderH00,boththeprimaryuserandmalicioususerareabsent,resultinginri=ni.It32followsthat:00=1NE8<:NXi=1nini9=;=1NNXi=1Efjnij2g=˙2n;(2.29)and˙200canbeobtainedas:˙200=Enj^Rrrj2oj00j2=1N2E8<:NXi=1NXj=1nininjnj9=;(˙2n)2=1N2"NXi=1Efjnij4g+NXi=1NXj=1j6=iEfjnij2gEfjnjj2g#(˙2n)2=1N2NEfj~nj4g+N(N1)(˙2n)2(˙2n)2=1NEfj~nj4g(˙2n)2;(2.30)whereweassumethatEfjnij4g=Efj~nj4g8i.Similarly,underH01,thereceivedsignalis33representedasri=mi+ni,andthemean01canbeobtainedasfollows:01=1NE8<:NXi=1(mi+ni)(mi+ni)9=;=1NE8<:NXi=1(mimi+nini)9=;=1N"NXi=1(Efjmij2g+Efjnij2g)#=˙2m+˙2n:(2.31)Thevariance˙201canbeobtainedas:˙201=Enj^Rrrj2oj01j2=1N2E8<:NXi=1(mi+ni)(mi+ni)NXj=1(mj+nj)(mj+nj)9=;(˙2m+˙2n)2=1N2"NXi=1(Efjmij4g+Efjnij4g+4Efjmij2gEfjnij2g+Ef2Ref(mi)2(ni)2gg)+NXi=1NXj=1j6=iEfjmij2gEfjmjj2g+Efjnij2gEfjnjj2g+2Efjmij2gEfjnij2g#(˙2m+˙2n)2=1NEfj~mj4g+Efj~nj4g+Ef2Ref(~m)2(~n)2gg+2˙2m˙2n(˙2m)2(˙2n)2;(2.32)whereweassumethatEfjmij4g=Efj~mj4gandEf2Ref(mi)2(ni)2gg=Ef2Ref(~m)2(~n)2gg8i.UnderH10,thereceivedsignalisexpressedasri=si+ni,andthemean10canbe34obtainedasfollows:10=1NE8<:NXi=1(si+ni)(si+ni)9=;=1NE8<:NXi=1(sisi+nini)9=;=1N"NXi=1(Efjsij2g+Efjnij2g)#=˙2s+˙2n;(2.33)and˙210canbeobtainedas:˙210=Enj^Rrrj2oj10j2=1N2E8<:NXi=1(si+ni)(si+ni)NXj=1(sj+nj)(sj+nj)9=;(˙2s+˙2n)2=1N2"NXi=1(Efjsij4g+Efjnij4g+4Efjsij2gEfjnij2g+Ef2Ref(si)2(ni)2gg)+NXi=1NXj=1j6=iEfjsij2gEfjsjj2g+Efjnij2gEfjnjj2g+2Efjsij2gEfjnij2g#(˙2s+˙2n)2=1NEfj~sj4g+Efj~nj4g+Ef2Ref(~s)2(~n)2gg+2˙2s˙2n(˙2s)2(˙2n)2:(2.34)Similarly,underH11,thereceivedsignalisrepresentedasri=si+mi+ni,andthemean3511canbeobtainedasfollows:11=1NE8<:NXi=1(si+mi+ni)(si+mi+ni)9=;=1NE8<:NXi=1(sisi+mimi+nini)9=;=1N24NXi=1(Efjsij2g+Efjmij2g+Efjnij2g)35=˙2s+˙2m+˙2n:(2.35)Thevariance˙211canbeobtainedas:˙211=Enj^Rrrj2oj11j2=1N2"NXi=1(Efjsij4g+Efjmij4g+Efjnij4g+4Efjsij2gEfjmij2g+4Efjsij2gEfjnij2g+4Efjmij2gEfjnij2g+Ef2Ref(si)2(mi)2gg+Ef2Ref(si)2(ni)2gg+Ef2Ref(mi)2(ni)2gg)+NXi=1NXj=1j6=iEfjsij2gEfjsjj2g+Efjmij2gEfjmjj2g+Efjnij2gEfjnjj2g+Efjsij2gEfjmjj2g+Efjsjj2gEfjmij2g+Efjsij2gEfjnjj2g+Efjsjj2gEfjnij2g+Efjmij2gEfjnjj2g+Efjmjj2gEfjnij2g#j11j2=1NEfj~sj4g+Efj~mj4g+Efj~nj4g+Ef2Ref(~s)2(~m)2gg+Ef2Ref(~s)2(~n)2gg+Ef2Ref(~m)2(~n)2gg+2˙2s˙2m+2˙2s˙2n+2˙2m˙2n(˙2s)2(˙2m)2(˙2n)2:(2.36)36Followingthediscussionsabove,wehave:~Pf;0=Prf^Rrr0jH00g=Q(000˙00);(2.37)and~Pf;1=Prf^Rrr1jH10g=Q(110˙10):(2.38)Similarly,wehave:~Pm;0=Prf^Rrr<0jH01g=1Q(001˙01);(2.39)and~Pm;1=Prf^Rrr<1jH11g=1Q(111˙11):(2.40)Theoverallfalsealarmrate~Pfandmissdetectionprobability~Pmcanbecalculatedfollowing(2.24),(2.27).Thatis:~Pf=P0Q(000˙00)+(1P0)Q(110˙10);(2.41)37and~Pm=1P0Q(001˙01)+(P01)Q(111˙11):(2.42)2.3.2.2TheOptimalThresholdsforMaliciousUserDetectionInthissubsection,weseektoobtaintheoptimalthresholds0;optand1;optthatminimizetheoverallmissdetectionprobabilityofthemaliciousnodedetectionproblem,whilemain-tainingthefalsealarmratesbelowacertainthreshold.Thisproblemcanbeformulatedasfollows:min~Pmsubjectto~Pf;0;and~Pf;1:(2.43)Itisnotedthattheproblemformulationaboveisequivalentto:min~Pm;0subjectto~Pf;0;(2.44)andmin~Pm;1subjectto~Pf;1:(2.45)Thus,werequest:~Pf;0=Q(000˙00);(2.46)38and~Pf;1=Q(110˙10);(2.47)whichimpliesthat:0˙00Q1()+00;(2.48)and1˙10Q1()+10:(2.49)Notethatinordertominimizetheoverallmissdetectionprobability~Pm,0in(2.48),and1in(2.49)shouldbeassmallaspossible.Hence,wesetthethresholdsto:0;opt=˙00Q1()+00;(2.50)and1;opt=˙10Q1()+10:(2.51)Bysubstituting0;optand1;optin(2.42),weobtaintheoverallmissdetectionprobabilityas:~Pm=1P0Q(˙00Q1()+0001˙01)+(P01)Q(˙10Q1()+1011˙11):(2.52)Proposition2.1Formalicioususerdetection,tominimizetheoverallmissdetectionproba-bility~Pmsubjecttothefalsealarmrateconstraints~Pf;0and~Pf;1,whichalsoensuresthat~Pf,weneedtochoose0;opt=˙00Q1()+00,and1;opt=˙10Q1()+10.392.4SecurityandFeasibilityoftheProposedAES-assistedDTVApproachThissectionisdevotedtodiscussthesecurityandfeasibilityoftheproposedAES-assistedDTVscheme.WebeginbyprovidingageneraloverviewoftheAESalgorithm.WethendiscussandinvestigatethesecurityandpracticabilityoftheAES-assistedDTVscheme.2.4.1ABriefOverviewoftheAESAlgorithmAdvancedEncryptionStandardisthecurrentNationalInstituteofStandardsandTech-nology(NIST)dataencryptionstandard,ithasbeenadoptedbytheU.S.DepartmentofCommercein2001aftergoingthroughalongevaluationperiod.Ithasbeenchosenbecauseofitssecurity(resistanceagainstallknownattacks),simplicity,availabilityintkeysizes,andinhardwareandsoftwareimplementations[54].AESisasymmetric-keycipher,inwhichasinglekeyisusedforbothencryptionanddecryption.Thekeyissharedbetweenthecommunicationparties,andkeptprivate.Figure2.3showsthegeneralstructureoftheAESencryptionalgorithm.Itmainlyconsistsoffourstagesthatareappliedtotheinputdata,whichisarrangedin44arrayofbytes.Thefourstagesarerepeated,andthenumberofrepetitionsdependsonthekeylength(128,192,or256bits).ThefourstagesofAESare:1.SubBytesStageInthisstage,eachbyteinthe44arrayissimplymappedtoanotherbytebasedonalookuptablecalledtheS-box.ThesecurityreasonforcreatingtheS-boxistothwartalltheknowncryptanalyticattacks[51].40Figure2.3:AESencryption.2.ShiftRowsStageHere,eachrowinthe44dataarray,excepttherow,isshiftedtotheleftbyanumberofbytes.Inparticular,thesecondrowisshiftedtotheleftby1byte,whilethethirdandfourthareshiftedby2bytesand3bytes,respectively.TheShiftRowsstageprovidesintheciphersothattheoutputoftheAESalgorithm(i.e.theciphertext)carriesnostatisticalrelationshiptotheinput(i.e.theplaintext)[51].3.MixColumnsInthisstage,eachbyteinacolumnisreplacedbyacombinationofthefoursbyteswithinthesamecolumn.TheMixColumnsoperationalsoprovidesproperty[51].4.AddRoundKeyInthisstage,eachbyteinthearrayisaddedtotheRoundKeyarrayusingbit-wiseXORfunction.TheAddRoundKeystageisusedtoimpacteverybitwithinthearray[51].412.4.2SecurityoftheAES-assistedDTVAsstatedearlier,AEShasbeenproventobesecureunderallknownattacks,inthesensethatitiscomputationallyinfeasibletobreakAESinrealtime.Inourcase,thismeansthatitiscomputationallyinfeasibleformalicioususerstoregeneratethereferencesignal.Moreover,theAESalgorithmhasaveryimportantsecurityfeatureknownastheavalanchect,whichmeansthatasmallchangeintheplaintextorthekeyyieldsalargechangeintheciphertext[51].Actually,evenifonebitischangedintheplaintext,theciphertextwillbechangedbyapproximately50%.Therefore,itisimpossibletorecovertheplaintextgiventheciphertextonly.Figure2.4:Normalizedcross-correlationbetweenthereferencesignalandnoisyversionsofmalicioususer'ssignal.Notethatthecross-correlationvaluesareintheorderof104,whichiscloseto0.ToillustratethesecurityoftheAES-assistedDTVbasedontheavalanchethecross-correlationbetweenthereferencesignalandmalicioussignalundertSNRvalues42Figure2.5:Normalizedcross-correlationbetweenthereferencesignalandnoisyversionsoftheprimaryuser'ssignal.Here,˙2s=1.isobtained,asshowninFigure2.4.Itcanbeseenthatthecross-correlationvaluesarearound0in(2.14),whichimpliesthatthemalicioussignalandthereferencesignalareuncorrelated.Ontheotherhand,thecross-correlationbetweenthereferencesignalandnoisyversionsoftheprimarysignalisshowntobeveryhigh(around1in(2.16)),underallSNRvalues,asdepictedinFigure2.5.ItshouldbeappreciatedthatintheDTVsystem,theminimumSNRis28:3dB[49].TheseresultsshowthattheAES-assistedDTVschemeissecureunderPUEA,asmali-cioususerscannotregeneratethereferencesignalinrealtime.2.4.3FeasibilityHere,weshowthatitispracticaltogeneratetherequiredsyncbitswithintheframetimedurationshowninFigure2.1.43TheAESalgorithmisoneoftheblockciphersthatcanbeimplementedintoperationalmodestogeneratestreamdata[55].High-throughput(3:84Gbpsandhigher)AESchipscanbefoundin[46,47].In[56],anexperimentwasperformedtomeasuretheAESalgorithmperformance,whereseveralsizesfrom100KBto50MBwereencryptedusingalaptopwith2:99GHzCPUand2GBRAM.Basedontheresultsoftheexperiment,whentheAESoperatesinthecipherfeedback(CFB)mode,554bytescanbeencryptedusing256-bitAESalgorithmin77:3.Therefore,eventhe2:99GHzCPUcangeneratetherequiredAESreferencesignalwithintheframetimeduration.NotethattheTVstationsgenerallyhavepowerfulprocessingunits,henceitisnotaproblemtogeneratetherequiredsecuresyncbitswithintheframeduration.With3:84Gbpsencryptionspeed,forexample,39KBcanbeencryptedin77:3,whichismorethanadequate.2.5SimulationResultsInthissection,wedemonstratetheivenessoftheAES-assistedDTVschemethroughsimulationexamples.First,weillustratetheimpactofthenoiselevelontheoptimalthresh-olds0;optand1;opt.Then,weevaluatethefalsealarmratesandmissdetectionprobabilitiesforbothprimaryuserandmalicioususerdetection.Inthesimulations,weassumethatsi,mi,andniarei.i.d.sequences,andareofzeromean.WefurtherassumethattheprimaryuserisabsentwithprobabilityP0=0:25.Theprimaryuser'ssignalpowerisassumedtobenormalizedto˙2s=1.Formalicioususerdetection,wesetthefalsealarmconstraint=103.Example1:Falsealarmrateandmissdetectionprobabilityforprimaryuserdetection.Using=˙2s=2,weobtainthefalsealarmrateandmissdetectionprobability44(a)ThefalsealarmratePf,thetwocurvesareiden-tical.(b)ThemissdetectionprobabilityPm,thetwocurvesareidentical.Figure2.6:Example1:Thefalsealarmrateandmissdetectionprobabilityforprimaryuserdetection.numericallyandcomparethemwiththetheoreticalresults.ThefalsealarmrateisillustratedinFigure2.6(a).ItisnotedthatthetheoreticalfalsealarmratePfin(2.18)dependson,since˙20isafunctionof.However,basedon(2.15)andtheavalancheoftheAESalgorithm,thisdependencybecomesnegligiblewhenNislarge.ThiscanbeseenfromFigure2.6(a)asthetheoreticalcalculationsmatchperfectlywiththenumericalsimulations.TheprobabilityofmissdetectionisshowninFigure2.6(b).Italsocanbeseenthatthetheoreticalcalculationsandnumericalsimulationsarematchedperfectly.ItisclearthattheproposedAES-assistedDTVapproachachieveszerofalsealarmrateandmissdetectionprobabilityunderalargerangeofSNRvalues.Example2:Theoptimalthresholdsformalicioususerdetection.Inthisexam-ple,wedemonstratetheoptimalthresholdsthatminimizethemissdetectionprobabilitiesunderanedconstraintonthefalsealarmratesformalicioususerdetection.Figure2.7showsthetwooptimalthresholds0;optand1;optversusSNRfor=103.WeobservethatthetwocurvesdecreaseastheSNRincreases,whichcanbevwith45Figure2.7:Example2:Theoptimalthresholdsformalicioususerdetectionfor=103.Here,P0=0:25.(2.50)and(2.51).Example3:Falsealarmrateandmissdetectionprobabilityformalicioususerdetection.Inthisexample,weobtaintheoverallfalsealarmrateandmissdetectionprobabilitynumericallyandcomparethemwiththetheoreticalresults.Figure2.8(a)showstheoverallfalsealarmrate~Pffor=103.Itisnotedthatthetheoreticalcalculationsandnumericalsimulationsarealmostequal,andthefalsealarmconstraintisTheoverallmissdetectionprobability~PmisillustratedinFigure2.8(b).ItisshownthattheproposedapproachachieveszerooverallmissdetectionprobabilityunderalargerangeofSNRvalues.46(a)Theoverallfalsealarmrate~Pf.(b)Theoverallmissdetectionprobability~Pm,thetwocurvesareidentical.Figure2.8:Example3:Theoverallfalsealarmrateandtheoverallmissdetectionprobabilityformalicioususerdetection.Here,P0=0:25and=103.2.6SummaryInthischapter,wepresentedtheproposedAES-assistedDTVschemeforrobustpri-maryandsecondarysystemoperationsunderprimaryuseremulationattacks.Inthepro-posedscheme,anAES-encryptedreferencesignalisgeneratedattheDTVtransmitterandusedasthesyncbitsoftheDTVdataframes.Byallowingasharedsecretbetweenthetrans-mitterandthereceiver,thereferencesignalcanberegeneratedatthereceiverandbeusedtoachieveaccuratefull-bandidenofauthorizedprimaryusers.Moreover,whencombinedwiththeanalysisontheauto-correlationofthereceivedsignal,thepresenceofthemalicioususercanbedetectedaccuratelynomattertheprimaryuserispresentornot.Sec-ond,weanalyzedthedetectionperformanceoftheproposedAES-assistedDTVapproach.Third,wediscussedthesecurityandfeasibilityoftheproposedAES-assistedDTVapproach.Throughsimulationexamples,itwasshownthattheproposedAES-assistedDTVschemecanachieveverylowfalsealarmratesandmissdetectionprobabilitieswhendetectingthe47primaryuserandmalicioususer.Thatis,withtheproposedAES-assistedDTVscheme,pri-maryuseremulationattackscanbeelycombated.Potentially,theproposedschemecanbeapplieddirectlytofutureIoTnetworksformoretspectrumsharingwiththeexistinglowerbandsystems.48Chapter3Sub-bandPUEADetectionandMitigationInthischapter,weconsidersub-bandPUEAdetectionintheOFDM-basedcognitiveradionetworks.TheDTVmodeladoptedisthesecondgenerationterrestrialdigitaltelevisionstandard(DVB-T2).WeproposeantAES-basedDTVscheme,wheretheexistingreferencesequenceusedtogeneratethepilotsymbolsintheDVB-T2framesisencryptedusingtheAESalgorithmforaccuratesub-bandprimaryuserandmalicioususerdetection.Withtheproposedscheme,wecandetectPUEAaccuratelyoverallsubcarriersorsub-bandswheretheP2symbolspresent.NumericalresultsshowthatwiththeAES-basedDTVscheme,boththeprimaryuserandmalicioususercanbedetectedaccuratelyunderprimaryuseremulationattacks.Potentially,theproposedschemecanbeapplieddirectlytofutureIoTnetworksformoretspectrumsharingwiththeexistinglowerbandsystems.3.1IntroductionCognitiveradionetworking[2{4]isapromisingtechniquetosolvethespectrumscarcityproblemandhasattractedalotofresearchinterestinrecentyears.ThebasicideaoftheCRnetworksistoidentifytheunoccupiedlicensedspectrumforsecondaryusers,without49interferingwiththeprimaryuser.AserioussecuritythreattotheCRnetworksisreferredtoasprimaryuseremulationattack[7{9],whereamalicioususeremulatesthesignalcharac-teristicsoftheprimaryuser,therebycausingtheSUstoerroneouslyidentifytheattackerastheprimaryuser.TheQoSforthesecondaryuserscanbeseverelydegradedbythepresenceofPUEAinCRnetworks[57].TodetectanddefendagainstPUEA,severalschemeshavebeenproposedinliterature.In[12],alocalization-basedtransmittervschemewasproposed.In[13]and[14],theauthorsproposedaRSS-basedbasedtechniquetodefendagainstPUEA,wheretheattackerscanbeidenbycomparingthereceivedsignalpowerlevelsoftheprimaryuserandthesuspectedattacker.In[18],theauthorsproposedcompressivesensingbasedapproachtodistinguishwhetherthesignalstransmittedarefromtheprimaryusersormali-cioususers.InmobileCRnetworks,theauthorsin[19]proposedamethodtodetectPUEAonmobileprimaryusersbyexploitingthecorrelationsbetweenradiofrequencysignalsandacousticinformationtoverifytheexistenceofmaliciouswirelessmicrophones.Recently,in[20],locationreliabilityandmaliciousintentionwereexploredtodetectprimaryuserandmalicioususerinmobileCRnetworks.Inmostoftheseapproaches,thedetectionofPUEAisbasedontheassumptionthat:whenthesignalisfromtheprimaryuser,thereceivedsignalstrengthatthesecondaryusersishigherthanthatfromamalicioususer.Asaresult,thesecondaryusercandistinguishtheprimarysignalfromthemalicioussignalbyexaminingthepowerlevelandthedirectionofarrivalofthereceivedsignal.Themajorlimitationwiththiskindofapproachesisthat:theymayfailwhenamalicioususerisatalocationwhereitproducesthesameDOAandcomparableRSSasthatoftheactualprimarytransmitter.Inthischapter,motivatedbytheprevalenceoftheOFDM-basedDTVstandard,weconsidersub-bandmalicioususerdetectioninOFDM-basedCRnetworkunderPUEA.Here,50Figure3.1:Theproposedcognitiveradionetworkarchitecture.theCRnetworkconsistsofprimaryusers,secondaryusercoordinators(SUCs),secondaryusers,andmalicioususers.TheSUCsaredesignedtoperformspectrumsensingandPUEAdetection,andthencoordinatetheSUsforcollision-freetransmissionoverthewhitespaces.WeproposearobustandtAES-basedDTVscheme,wheretheexistingreferencesequenceusedtogeneratethepilotsymbolsintheDVB-T2framesisencryptedusingtheAESalgorithm,andtheresultedsequenceisexploitedforaccuratedetectionoftheauthorizedprimaryuserandthemalicioususer.3.2SystemModelWeconsideranOFDM-basedcognitiveradionetworkunderPUEA,asshowninFigure3.1.Inthismodel,weassumethatthenetworkisdividedintohexagonallyshapedareas(cells),eachofwhichconsistsofasecondaryusercoordinatorlocatedatthecenterofthe51Figure3.2:TheDVB-T2framestructure.cell,asetofsecondaryusers,andsomemalicioususers.Furthermore,thenetworkincludespowerfulprimaryusers(DTVtransmitters)distributeduniformlythroughoutthenetwork.TheprimaryusersareassumedtobeusingOFDMschemefordatatransmission,whichisthecaseinmostoftheexistingorfutureDTVsystems[58].Here,weconsiderthesecondgenerationterrestrialdigitaltelevisionstandard(DVB-T2)systemasanexampleduetoitsprevalenceand.TheframestructureoftheDVB-T2standardisshowninFigure3.2[59].Itconsistsofsuperframes,whicharepartitionedintoT2-framesandsupplementaryfutureextensionframes(FEF).EachT2-framehasthreekindsofOFDMsymbols:P1preamblesymbolusedforcharacterizingthebasictransmissionparameters,P2preamblesymbol(s)usedforcarryingsignalinginformation,anddatasymbolsforpayload.Furthermore,eachsymbolhastpilotsymbolsusedforframesynchronization,frequencysynchronization,andchannelestimation[59].Here,weproposetousetheP2pilotsforthedetectionoftheprimaryuserandmalicioususerfortworeasons.First,theyaretheonlypilotswhosefrequencylocationsareindependentoftheFFTsize(1K,2K,4K,8K,16K,32K)andthe52operationalmodes(SISO,MIMO)exceptin32KSISOmode.Second,theyhavethelargestnumberamongallthepilotsymbols,whichcanbeexploitedtoachieveedetectionoftheauthorizedprimaryuserandthemalicioususer.WewouldliketopointoutthatwiththeOFDMstructureusedinDVB-T2,theproposedschemecanaccuratelydetectthepresenceofthemalicioususeroverallthesubcarrierswheretheP2pilotspresent.NotethattheP2pilotspresentonsubcarrierswhoseindexesareintegermultiplesof3,whichimpliesthatwecandetectthepresenceofthemalicioususersovereach3-subcarriersub-band.However,theproposedapproachcanbeusedtodetectPUEAovereachsinglesubcarrierifwecanspreadthepreambleP2symbolsintheDVB-T2standardtocovereverysubcarrier,insteadofputtingmultipleP2symbolsovereachselectedsubcarrier.Thedeploymentofthesecondaryusercoordinatorsintheproposednetworkarchitecturecanprovidetspectrumsharingbetweentheprimaryuserandthesecondaryusers,andalsoamongthesecondaryusers.Moresp,thesecondaryusercoordinatorcandetectidlespectralspacesinthefrequencyspectrum,andassignthesespacestothesecondaryuserswithinthecell.Withasecurereferencesignalsharedbetweentheprimaryuserandthesecondaryusercoordinator,thesecondaryusercoordinatorcanalsodetectstheexistenceofthemalicioususer.Itshouldalsobeemphasizedthatwhenasecondaryusercoordinatorisinvolved,onlythesecondaryusercoordinatorneedstoperformprimaryuserandmalicioususerdetection.Hence,itcanreducetheburdenofeachindividualsecondaryuser,whichgenerallyreliesontransceiverswithlimitedcapabilities[60].533.3AES-basedPUEADetectionSchemeInthissection,wepresenttheAES-basedPUEAdetectionschemeforreliableandectiveCRnetworkoperation.WestartbydiscussingtheprimaryDTVtransmitterdesign,wheretheexistingreferencesequenceusedtogeneratetheP2pilotsisencryptedusingtheAESalgorithm.Then,weinvestigatetheSUCreceiverdesignforaccuratedetectionoftheprimaryuserandmalicioususer.Moreover,weevaluatethedetectionperformancethroughfalsealarmrateandmissdetectionprobability,anddiscusstheofthesamplesizeonthedetectionmeasures.3.3.1PrimaryUserTransmitterDesignThefrequencyindexesofP2pilotsaredeterminedby:=fmjm0(mod3);0mMmaxg;(3.1)whereMmaxisthemaximumfrequencyindexfortheP2symbol.TheP2pilotsaregeneratedbasedonthereferencesequencers,whichisobtainedbyperformingtheXOR(exclusive-OR)functiontotwopseudo-randomsequences,asshowninFigure3.3[59].MoredetailsonthePNsequencesgenerationcanbefoundin[59].AsillustratedinFigure3.3,weproposethattheexistingreferencesequencersisfurtherencryptedusingtheAESalgorithmwithasecretkeytoobtaintheproposedreferencesignalsasfollows:s=E(key;rs);(3.2)whereE(;)denotestheAESencryptionoperation.Itshouldbenotedthattheencrypted54Figure3.3:Generationoftheproposedreferencesignals.pilotsymbolsarestillusedastheconventionalpilotsforsynchronizationpurposes.ThesecretkeycanbegeneratedanddistributedtotheDTVtransmitterandreceiverfromatrusted3rdpartyinadditiontotheDTVandtheSUCs.The3rdpartyservesastheauthenticationcenterforboththeprimaryuserandtheCRuser,andcancarryoutkeydistribution.Topreventimpersonationandreplayattacks,thekeyshouldbetimevarying[51,61].Inthecognitiveradionetwork,authenticationandauthorizationareneededbeforekeydistribution,toensurethatonlyauthorizeduserswouldreceivethekey,andtoenforceuseraccountabilityandpreventmisuseofprivileges.3.3.2SecondaryUserCoordinatorReceiverDesignTheSUCreceiverregeneratestheencryptedreferencesignalwiththesecretkeysharedbe-tweentheprimarytransmitterandthereceiver.Acorrelationdetectorisemployed,whereforprimaryuserdetection,thereceiverevaluatesthecross-correlationbetweenthereceivedsignalandtheregeneratedreferencesignal;formalicioususerdetection,thereceiverfurtherevaluatestheauto-correlationofthereceivedsignal.Toperformprimaryuserandmali-cioususerdetection,theSUCreceiverneedstocollectatamountofthereceivedsamples,Msamples(seeSection3.3.3),onthedesiredsubcarrierbeforecalculatingthecross-correlationandtheauto-correlation.55Foranyk2=f0;3;6;:::;Mmaxg,thereceivedM1vectoronsubcarrierkatSUC1,showninFigure3.1,canbemodeledas:rk=kqld1Ps;kjhkjsk+kmk+nk;(3.3)wheresk,Ps;k,ld1,andhksCN(0;1)arethetransmittedbinaryphaseshiftkeying(BPSK)symbolvector(BPSKisadoptedherebecauseitisusedintheDVB-T2standard),theallocatedtransmitpower,thepathlossattenuationfactor,andtheRayleighfadingcoetonsubcarrierkbetweenPU1andSUC1,respectively.mkistheMU1BPSKsymbolvector,nksN(0;˙2nI)isthenoise,kandkarebinaryindicatorsforthepresenceoftheprimaryuserandmalicioususeronk2subcarrier,respectively.3.3.2.1PrimaryUserDetectionTodetectthepresenceoftheprimaryuseronsubcarrierk,wherek2theSUCreceiverevaluatesthecross-correlationbetweenthereceivedvectorandtheregeneratedreferencevector.Thenormalizedcross-correlationcanberepresentedas:Rrs;k=<rk;sk>M=kqld1Ps;kjhkj;(3.4)whereweassumethatthechannelistime-invariantduringtheobservedperiod,andthechannelgainisavailableattheSUCreceiver,andsk,mk,nkareindependentwitheachotherandareofzeromean.Dependingonthevalueofk,theSUCreceiverdecideswhethertheprimaryuserispresentorabsentonsubcarrierk,andthenpassestheinformationdetectiontotheSUswithinthecell.56Assumingthatthesignalsareergodic,thentheensembleaveragecanbeapproximatedbythetimeaverage.Here,weusethetimeaveragetoestimatethecross-correlation,i.e.,^Rrs;k,1MMXi=1rk;isk;i;(3.5)whereMisthesamplesize,sk;iandrk;idenotetheithentriesinthereferenceandreceivedvectorsonsubcarrierk2respectively.Todetectthepresenceoftheprimaryuser,thereceivercomparesthecross-correlationbetweenthereferencesignalandthereceivedsignalwithathresholdTk.Wehavetwocases:(i)if^Rrs;kTk,thenwedeterminethattheprimaryuserispresent,and(ii)if^Rrs;k<Tk,thenwedeterminethattheprimaryuserisabsent.Thedetectionproblemcanbemodeledasabinaryhypothesistestproblemasfollows:H0:theprimaryuserisabsentonsubcarrierk(k=0)H1:theprimaryuserispresentonsubcarrierk(k=1)Ascanbeseenfrom(3.4),thenormalizedcross-correlationbetweenthereferencesignalandthereceivedsignalisequalto0orpld1Ps;kjhkj,incasewhentheprimaryuserisabsentorpresent,respectively.Followingtheminimumdistancerule,wechooseTk=(pld1Ps;kjhkj)=2asthethresholdforprimaryuserdetection.3.3.2.2MaliciousUserDetectionFormalicioususerdetection,theSUCreceiverfurtherevaluatestheauto-correlationofthereceivedvectorrk,wherethenormalizedauto-correlationcanberepresentedas:Rrr;k=<rk;rk>M=kld1Ps;kjhkj2+k˙2m;k+˙2n:(3.6)57Basedontheestimatedvalueofk,kcanbedeterminedaccordinglythrough(3.6).Wehavethefollowingcases:Rrr;k=8>>>>>>>>><>>>>>>>>>:ld1Ps;kjhkj2+˙2m;k+˙2n;k=1;k=1ld1Ps;kjhkj2+˙2n;k=1;k=0˙2m;k+˙2n;k=0;k=1˙2n;k=0;k=0(3.7)Assumingergodicsignals,wecanusethetimeaveragetoestimatetheauto-correlationasfollows:^Rrr;k,1MMXi=1rk;irk;i:(3.8)Thresholdbaseddetectionmethodcanbedevelopedaccordingly.Here,wecanmodelthedetectionproblemusingfourhypotheses,denotedbyHkk,wherek;k2f0;1g:H00:themalicioususerisabsentgiventhatk=0(^Rrr;k<Tk;0)H01:themalicioususerispresentgiventhatk=0(^Rrr;kTk;0)H10:themalicioususerisabsentgiventhatk=1(^Rrr;k<Tk;1)H11:themalicioususerispresentgiventhatk=1(^Rrr;kTk;1)Theperformanceofthedetectionprocessfortheprimaryuserandmalicioususerisevaluatedthroughthefalsealarmrates(FARs)andthemissdetectionprobabilities(MDPs),aswillbediscussedinthefollowingsubsection.583.3.3FalseAlarmRateandMissDetectionProbabilityDenotethefalsealarmrateforprimaryuserdetectiononsubcarrierk,wherek2byPf;k,andthemissdetectionprobabilitybyPm;k.Thefalsealarmrateisdastheconditionalprobabilitythattheprimaryuserisconsideredtobepresent,whenitisactuallyabsent,i.e.,Pf;k=Pr(H1jH0);(3.9)whereasthemissdetectionprobabilityisastheconditionalprobabilitythattheprimaryisconsideredtobeabsent,whenitispresent,i.e.,Pm;k=Pr(H0jH1):(3.10)Notethatthecross-correlation^Rrs;k,in(3.5),isthesamplemeanofindependentrandomvariablesofsizeM.Accordingtothecentrallimittheorem(CLT),aslongasMistlylarge(i.e.,M30),thissamplemeanapproximatelyfollowstheGaussiandistribution[62].Moresp,underH0,^Rrs;k˘N(0;˙20)withmean0=0andvariance˙20=1Mk˙2m;k+˙2n.Similarly,underH1,^Rrs;k˘N(1;˙21)withmean1=pld1Ps;kjhkjandvariance˙21=1Mld1Ps;kjhkj2Efj~skj4g+k˙2m;k+˙2nld1Ps;kjhkj2,whereEfjsk;ij4g=Efj~skj4g8i.ThefalsealarmratePf;kcanbeobtainedas:Pf;k=Prf^Rrs;kTkjH0g=Q(Tk0˙0);(3.11)whereQ(x)=1p2ˇR1xeu22dudenotesthetailprobabilityofthestandardnormaldistribu-59tion.ThemissdetectionprobabilityPm;kcanbeobtainedas:Pm;k=Prf^Rrs;k<TkjH1g=1Q(Tk1˙1):(3.12)Similarly,formalicioususerdetection,denotethefalsealarmrateby~Pf;kandthemissdetectionprobabilityby~Pm;k,whichcanbeobtainedas:~Pf;k=P0Q(Tk;000˙00)+(1P0)Q(Tk;110˙10);(3.13)and~Pm;k=1P0Q(Tk;001˙01)+(P01)Q(Tk;111˙11);(3.14)whereP0=Pr(k=0),and00,˙00,01,˙01,10,˙10,11,˙11,andtheoptimalvaluesforTk;0andTk;1canbefoundintheappendix.3.3.4DetectionPerformanceVersusSampleSizeInthissubsection,wediscusstheofthesamplesizeMonthethefalsealarmrateandmissdetectionprobability.Substitutingby(pld1Ps;kjhkj)=2forTkin(3.11)and(3.12),wehave:Pf;k=Q(cpM);(3.15)andPm;k=1Q(cpM);(3.16)wherec=(pld1Ps;kjhkj)=(2qk˙2m;k+˙2n).60Discussions:Following(3.15)and(3.16),underreasonablyhighSNRlevel,boththefalsealarmratePf;kandthemissdetectionPm;kdecreaseasthesamplesizeincreases.Moresp,limM!1Pm;k=0andlimM!1Pf;k=0.3.4SimulationResultsInthissection,wedemonstratetheperformanceoftheproposedAES-basedDTVschemethroughtheevaluationofthefalsealarmrateandmissdetectionprobabilityforprimaryuserandmalicioususerdetection.Inthefollowing,weassumethatsk,mk,nkareindependentwitheachotherandareofzeromean.Furthermore,thepathlossattenuationfactorisgivenbyld1=d1,where=2isthepathlossexponentandd1=500misthedistancesbetweenPU1andSUC1,showninFigure3.1.Figure3.4:Thefalsealarmrate(FAR)andmissdetectionprobability(MDP)versusSNRforprimaryuserdetection.61Figure3.5:Thefalsealarmrate(FAR)andmissdetectionprobability(MDP)versusSNRformalicioususerdetection.UsingthesamplesizeM=100,weobtainthefalsealarmrateandmissdetectionprobabilityforbothprimaryuserdetectionandmalicioususerdetection.Figure3.4showsthefalsealarmrateandmissdetectionforprimaryuserdetectionversusSNR,whilethefalsealarmrateandmissdetectionformalicioususerdetectionaredepictedinFigure3.5.ItcanbeseenthattheproposedAES-basedDTVschemecanachieveverylowfalsealarmratesandmissdetectionprobabilitiesevenunderverylowSNRvalueswhendetectingtheprimaryuserandmalicioususer.3.5SummaryInthischapter,westudiedthePUEAproblemintheOFDM-basedCRnetworks,andproposedareliableandtAES-basedDTVscheme.Intheproposedapproach,the62existingreferencesequenceintheDTVsystemisencryptedusingtheAESalgorithmforaccuratesub-banddetectionoftheauthorizedprimaryuser,aswellasmalicioususer,intheallocatedfrequencyspectrum.Itwasshownthat,withtheAES-basedDTVscheme,boththeprimaryuserandthemalicioususercanbedetectedaccuratelyunderprimaryuseremulationattacks.Theproposedapproachispracticallyfeasibleinthesensethatitcanelydetectprimaryuserandmalicioususerwithnochangeinhardwareorsystemstructureexceptofaplug-inAESchip.Potentially,theproposedschemecanbeapplieddirectlytofutureIoTnetworksformoretspectrumsharingwiththeexistinglowerbandsystems.63Chapter4EnergyHarvestingandReliableInformationTransferThischapteraimstostudysimultaneouswirelessinformationtransferandenergyharvestinginOFDM-basedCRnetworksunderPUEA.Here,weproposeanecommunica-tionschemeforthesecondaryusersunderPUEAbyexploitingtheenergyharvestingtech-niques.Optimalpowersplittingisconsideredforsum-ratemaximization.Astheoptimalsolutionreliesonmulti-dimensionalexhaustivesearch,weproposeanesuboptimalsolutionwithmuchlowercomputationalcomplexity.Itisobservedthatthesum-rateofthesecondaryusernetworkcanbeimprovedtlywiththeenergyharvestingtechnique.Then,weevaluatetheworst-casePUEAinterferenceintermsofminimizingthesum-rateforthesecondaryusers.Weshowthatforpracticalsystems,asCSIamongtheSUsisnotavailabletothemalicioususer,theworstjammingfortheSUsiswhenthemalicioususerperformsequalpowerallocationoverallthewhitespacesubcarriers.Simulationresultsareprovidedtoillustratetheproposedapproaches.Inadditiontothejammingscenario,energyharvestingcanbeappliedinthebenignenvironmentaswelltoresolvetheenergyscarcityintheIoTnetworks.644.1IntroductionAlongwiththeemergingenergy-constrainedIoTnetwork,whichbasicallyreliesontransceiverswithlimitedpowercapabilities,energyharvestingfromambientRFsignalshasattractedmoreresearchattentioninrecentyears[24{35].Theconceptofsimultaneouswirelessinformationtransferandenergyharvestingwasproposedin[24],wherethefundamentalbetweeninformationreceivingandpowertransferwasstudiedfromaninformationtheoryperspective.Inliterature[25,26],therehavebeentworepresenta-tivereceiverarchitecturesforenergyharvesting:power-splittingandtime-switching.Inthepower-splittingbasedapproach,apowersplitterisemployedattheRFbandatthereceiver.ThereceivedRFsignalissplitintotwoparts:oneforenergyharvestingandtheotherforin-formationprocessing.Inthetime-switchingapproach,thereceivedsignalisdividedintotwopartsinthetimedomain.Thepowertransferofenergyharvestingtechniquesde-pendsmainlyonthereceivedsignalstrength.In[27],itwasshownthatanof0:4%,18:2%andabove50%canbeachievedwithRFenergytransferwithinputpower40dBm,5dBm,and20dBm,respectively.Moredetailsonperformanceofwirelessenergytransfertechniquescanbefoundin[28].Recently,energyharvestinghasbeenwidelydiscussedinOFDM-basedsystems[29],wirelessrelaynetworks[30],andcognitiveradionetworks[31{35].Forexample,theauthorsin[31]investigatedtheoptimaltransmissionandenergyharvestingpolicytomaximizethesecondarynetworkthroughputinacognitiveradionetwork,wherethesecondaryusersop-portunisticallyaccessthespectrumorharvestenergyfromtheambientRFsignalsfromprimaryusers.In[32],theauthorsstudiedthechannelaccessprobleminCRnetworksinwhichasecondaryusercanchoosetoaccessthechannelfordatatransmissionorenergyhar-65vesting.Toobtainthechannelaccess,theypresentedanoptimizationformulationbasedonMarkovdecisionprocess.Generally,PUEAwasnotconsideredintheseworks[31{35].How-ever,PUEAisessentiallyjamminginterferencefortheSUs,andpotentiallytheperformanceofSUscanbeimprovedtlybyexploitingPUEAasanextraenergyresource.Inad-ditiontothejammingscenario,energyharvestingcanbeappliedinthebenignenvironmentaswelltoresolvetheenergyscarcityintheIoTnetworks.InChapter3,weinvestigatedsub-bandPUEAdetectionbyproposingarobustAES-basedDTVscheme,wheretheexistingreferencesequenceusedtogeneratethepilotsymbolsintheDVB-T2framesisencryptedusingtheAESalgorithm,andtheresultedsequenceisexploitedforaccuratesub-banddetectionoftheauthorizedprimaryuserandthemalicioususer.TheSUCsweredesignedtoperformspectrumsensingandPUEAdetection,andthencoordinatetheSUsforcollision-freetransmissionoverthewhitespaces.Inthischapter,weproposeanenergyharvestingschemeasfollows.WhenthedetectionoftheprimaryuserandthemalicioususerisdoneattheSUC,theSUCwillcoordinatethetransmissionsandspectrumsharingoftheSUsbasedonthedetectionresults.Iftherearenowhitespacesdetected(i.e.,thePUisoccupyingalltheallocatedspectrum),thensecondaryusersperformenergyharvestingfromthereceivedPUandMUsignals.Otherwise(i.e.,ifthereexistsomewhitespaces),secondaryusersharvestenergyandtransmittheirownsignalssimultaneouslyovertheassignedtimeslotsandfrequencybands.4.2TheProposedEnergyHarvestingSchemeWeconsideratypicalcell,whichconsistsofaprimaryuser,asecondaryusercoordinator,amalicioususer,andamultiple(K)ofsecondaryusers,andproposeareliableand66Figure4.1:Theproposedenergyharvestingscheme.cientschemeforresourceutilizationbetweenthesecondaryusers,showninFigure4.1.Inthisscheme,weproposethatSUC1detectsPU1andMU1usingcorrelationdetectorsafterobtainingMsamplesofthereceivedsignalonthedesiredsub-channel(seeSection3.3).Duringtheinitialdetectionperiod,alltheSUsharvestenergyfromallthesubcarriers.ThedetectionresultisthenusedbySUC1tocoordinatetransmissionsamongtheSUs.Assumingthereexistsomewhitespaces,thentheSUssharethesewhitespacesusingacollision-freemultipleaccessscheme.Withoutlossofgenerality,here,wechooseTimeDivisionMultiplex-ing(TDM),asshowninFigure4.1.Moresp,fortheM1=bM=Kcsamples,SU1transmitsitssignal,whiletheremainingSUsprocessinformationandharvestenergyfromthereceivedsignalswithapowersplittingreceiver[63].ForthenextM1samples,SU2transmitswithitsinherentpowerandtheharvestedpower,whiletheremainingSUsperforminformationrecoveryandenergyharvesting,andsoon.FortheremainingM2=MKM167samples,weproposethatalltheSUsharvestenergyfromthereceivedsignals,includingboththeprimaryuser'ssignalandthemalicioususer'ssignal.Notethatwhentheallo-catedchannelbandwidthisfullyutilizedbytheprimaryuser,theharvestedpower,bythesecondaryusers,canbestoredininternaland/orexternalbatteriesforlaterusages.4.3EnergyHarvestingandPerformanceOptimizationInthissection,wediscussasecondaryusersenergyharvestingandinformationexchangetechniquetoimprovethesystemperformanceoftheOFDM-basedCRnetworksunderPUEA.Toillustratetheproposedapproaches,wediscusstheoptimaltransceiverdesignformaximizingthesum-ratebetweentwosecondaryusersunderjamming.4.3.1SecondaryUsersEnergyHarvestingandInformationRe-coveryAsmentionedearlier,iftheprimaryuserispresentonallthesubcarriers,thenSUsonlyharvestenergyfromthereceivedsignalssincetheyareprohibitedfromtransmittingaslongastheprimaryuseristransmitting.Ontheotherhand,ifthereexistsomeidlespectralspaces,thenSUsprocessinformationandharvestenergysimultaneouslyfromthereceivedRFsignals,basedontheproposedmethodshowninFigure4.1.Here,weconsidertwo-waycommunicationbetweenSU1andSU2.First,weinvestigatethedownlinktransmission,wherefortheM1samples,SU1transmitstoSU2,meanwhileSU2performsinformationrecoveryandenergyharvestingfromthereceivedsignalswithapowersplittingreceiver.Then,forthenextM1samples,weconsidertheuplinktransmissionsimilarlyasthedownlinktransmissioncase,exceptthatthetransmissionthistimeisfromSU2toSU1.Forthe68remainingM2=M2M1samples,bothSU1andSU2harvestenergyfromthereceivedsignals,includingboththeprimaryuser'ssignalandthemalicioususer'ssignal.InthecasethattheSUshavetpower,formaximaltransmissionrate,wechooseM1=M=2.4.3.1.1AchievableTransmissionRateforSU1ConsideranOFDMsystemwithNsubcarriers.Then,thedownlinktransmissioncorre-spondingtothereceivedM1samplesonsubcarriern2Nsc=f1;2;:::;Ng,atSU2,canbemodeledas:r2;n=npld2Ps;ng2;n~sn+1;nqld4P1;nf1;n~s1;n+n~m2;n+~na;n;(4.1)wherePs;nisthePU1'stransmitpower,andld2,g2;nsCN(0;1),~sndenotethepathlossattenuationfactorfromPU1toSU2,theRayleighfadingcotbetweenPU1andSU2,andthePU1'stransmittedsymbolM11vectoronthenthsubcarrier,respectively;ld4,P1;n,f1;nsCN(0;1),and~s1;ndenotethepathlossattenuationfactorfromSU1toSU2,SU1'stransmitpower,theRayleighfadingcotbetweenSU1andSU2,andSU1'stransmittedsymbolvectoronthenthsubcarrier,respectively;~m2;nisthereceivedjammingsymbolvectorfromMU1,~na;nsCN(0;˙2aI)isthenoise,n,n,1;narebinaryindicatorsforthepresenceoftheprimaryuser,malicioususer,andSU1onthenthsubcarrier,respectively.Notethat1;n=1,forn2A1and1;n=0,forn=2A1,whereA1isthesetofwhitespacesubcarriersassignedtoSU1.LetA=fkjk=0;k2gbethesetofallidlesubcarriers.Withoutlossofgenerality,sinceSU1andSU2sharethewhitespacesusingtimedivisionmultipleaccess(TDMA),weassumethatA1=A.Thereceivedsignal,atSU2,issplitintotwoparts:oneforenergyharvestingandthe69otherforinformationdecoding.Letˆ22[0;1]denotethepowersplittingratioatSU2,thatis,pˆ2r2;nisusedforenergyharvesting.Then,from(4.1),thetotalharvestedpoweratSU2canbeobtainedas:PEH;2=ˆ2M1NXn=1(nld2Ps;njg2;nj2+1;nld4P1;njf1;nj2+n~˙22;m;n)+N˙2a;(4.2)where0<<1istheenergyconversion.Theotherpartofthereceivedsignal,i.e.p1ˆ2r2;n,isthenusedforinformationdetection.Thebasebandsignalcanbeexpressedas:~r2;n=q(1ˆ2)ld4P1;nf1;n~s1;n+p1ˆ2(n~m2;n+~na;n)+~nb;n;n2A;(4.3)where~nb;nsCN(0;˙2bI)isthenoiseinthebaseband.Notethatthereisnoprimaryusersignalinthewhitespaces.From(4.3),theachievablerateforthelinkfromSU1toSU2isgivenby:R1=M1MXn2Alog2 1+(1ˆ2)ld4P1;njf1;nj2(1ˆ2)(n~˙22;m;n+˙2a)+˙2b!:(4.4)Remark4.1Wewouldliketopointoutthat,theoretically,weshouldharvestalltheen-ergy(i.e.,chooseˆ2=1)overthebandoccupiedbythePU,useenergyharvesting(i.e.,chooseˆ2=0)forinformationrecoveryoverjamming-freewhitespaces,andperformpowersplittingonlyoverthewhitespaceswhicharecontaminatedbythemalicioususer.However,inpracticalsystems,onereceiveantennaisusedforallthesubcarriers.Asaresult,thesamepowersplittingratiohastobeappliedforallthebands,asshownin(4.2).Thesameargumentappliestothenextsubsection.704.3.1.2AchievableTransmissionRateforSU2TheuplinktransmissionfromSU2toSU1correspondingtothenextreceivedM1samplesonsubcarriern2Nsc=f1;2;:::;Ng,atSU1,canbemodeledas:r1;n=npld3Ps;ng1;n~sn+2;nqld4P2;nf2;n~s2;n+n~m1;n+~na;n;(4.5)wherePs;nisthePU1'stransmitpower,andld3,g1;nsCN(0;1),~sndenotethepathlossattenuationfactorfromPU1toSU1,theRayleighfadingcotbetweenPU1andSU1,andPU1'stransmittedsymbolM11vectoronthenthsubcarrier,respectively;P2;n,f2;nsCN(0;1),and~s2;ndenoteSU2'stransmitpower,theRayleighfadingcotbetweenSU2andSU1,andSU2'stransmittedsymbolvectoronthenthsubcarrier,respectively;~m1;nisthereceivedjammingsymbolvectorfromMU1,n,n,2;narebinaryindicatorsforthepresenceoftheprimaryuser,malicioususer,andSU2onthenthsubcarrier,respectively.Notethat2;n=1,forn2A2and2;n=0,forn=2A2,whereA2isthesetofwhitespacesubcarriersassignedtoSU2.Here,sinceSU1andSU2sharethewhitespacesusingTDMA,weassumethatA1=A2=A.Thereceivedsignal,atSU1,isthensplitintotwoparts:oneforenergyharvestingandtheotherforinformationrecovering.Letˆ12[0;1]denotethepowersplittingratioatSU1,thatis,pˆ1r1;nisusedforenergyharvesting.Then,from(4.5),thetotalharvestedpoweratSU1canbeobtainedas:PEH;1=ˆ1M1NXn=1(nld3Ps;njg1;nj2+2;nld4P2;njf2;nj2+n~˙21;m;n)+N˙2a:(4.6)Theotherpartofthereceivedsignal,p1ˆ1r1;n,isthenusedforinformationdetection.71Thebasebandsignalcanbeexpressedas:~r1;n=q(1ˆ1)ld4P2;nf2;n~s2;n+p1ˆ1(n~m1;n+~na;n)+~nb;n;n2A;(4.7)where~nb;nsCN(0;˙2bI)isthenoiseinthebaseband.From(4.7),theachievableratefortheuplinktransmissionisgivenby:R2=M1MXn2Alog2 1+(1ˆ1)ld4P2;njf2;nj2(1ˆ1)(n~˙21;m;n+˙2a)+˙2b!:(4.8)Discussions:IfM1<M=2,i.e.,M2=M2M1>0,thenbothSU1andSU2harvestenergyfromthereceivedsignalsduringtheM2period.Therefore,thetotalharvestedpowerattheSU1andSU2duringtheM2periodcanbeobtainedas:~PEH;1=(M2M1)NXn=1(nld3Ps;njg1;nj2+n~˙21;m;n)+N˙2a;(4.9)and~PEH;2=(M2M1)NXn=1(nld2Ps;njg2;nj2+n~˙22;m;n)+N˙2a:(4.10)InthecasethattheSUshavetpower,thenformaximaltransmissionrate,wechooseM1=M=2.4.3.2Sum-rateOptimizationInthissubsection,weinvestigatetheoptimaltransceiverdesignthataimstomaximizesum-ratebetweenSU1andSU2.72From(4.4)and(4.8),thesum-ratemaximizationproblemcanbeexpressedas:maxˆ1;ˆ2;M1;P1;P2Rsum=R1+R2s.t.0ˆ11;0ˆ21;0M1M2;Xn2AP1;nPSU1+PEH;1+~PEH;1M1;Xn2AP2;nPSU2+PEH;2+~PEH;2M1;(4.11)wherePj=fPj;n;8n2Ag;j=1;2,andPSU1,PSU2denotetheinherentpowersupplyofSU1andSU2,respectively.Theoptimizationproblemin(4.11)belongstotheclassofnon-convexoptimizationprob-lemsduetothepresenceofˆ1,ˆ2,andM1.However,forˆ1,ˆ2,andM1,theharvestedpowersin(4.2)and(4.6)arebothlinearfunctionsofP1andP2.Hence,theconstraintsin(4.11)becomelinearconstraints.Basedontheseobservations,itcanbeseenthattheproblemin(4.11)turnsouttobeconvexwhenˆ1,ˆ2,andM1areandcanbesolvednumericallyusingthestandardalgorithmsforconvexoptimization[64].Hence,thesum-ratemaximizationproblemin(4.11)fortheproposedschemecanbesolvedbythree-dimensionalexhaustivesearchoverˆ1,ˆ2,andM1.Notethatalthoughexhaustivesearchoverˆ1,ˆ2,andM1alongwiththeconvexop-timizationtechniquesguarantiesthegloballyoptimalsolutionof(4.11),itrequireshighcomputationalcomplexity.Toreducethecomplexity,weproposeasuboptimalsolution.Discussions:Thesum-rateoptimizationproblemcanbeextendedtotheK-usercaseby73dividingtheSUsintogroupsoftwo,thenperformingthesamesum-rateoptimizationasin(4.11).IfKisodd,thentheSUscanbedividedintoK12pairsplus1SU.Thesum-rateoptimizationforthelastSUisactuallymuchsimpler.Therefore,theoverallsum-rateformultipleuserswouldbesimilartothatofthetwo-usercase.Takingthelargenumberofwhitespacesincognitiveradios,itwouldbebtokeepthenumberofpairgroupssmallineachshared/assignedfrequencyband.4.3.3SuboptimalSolutionInthissubsection,wecalculatethesum-ratelowerbound,andthenderivethesubop-timalsolutionbasedonit.4.3.3.1Sum-rateLowerBoundFormaximaltransmissionrate,letM1beasM1=M=2andconsiderequalpowerallocationatthetwoSUs,i.e.,P1;n=1jAjPSU1+PEH;1M1;(4.12)andP2;n=1jAjPSU2+PEH;2M1:(4.13)74Then,theachievablerateforSU1canbelowerboundedby:R1=M1MXn2Alog2 1+(1ˆ2)ld4P1;njf1;nj2(1ˆ2)(n~˙22;m;n+˙2a)+˙2b!>M1MXn2Alog2 (1ˆ2)ld4P1;njf1;nj2(1ˆ2)(n~˙22;m;n+˙2a)+˙2b!M1MXn2Alog2 (1ˆ2)ld4P1;njf1;nj2(1ˆ2)~˙22;m+˙2b!;(4.14)where~˙22;m=1jAjPn2An~˙22;m;n+˙2a.Intheinequalityabove,wehaveusedthefactthatQNn=1xn1NPNn=1xnN.Similarly,theachievablerateofSU2islowerboundedby:R2M1MXn2Alog2 (1ˆ1)ld4P2;njf2;nj2(1ˆ1)~˙21;m+˙2b!;(4.15)where~˙21;m=1jAjPn2An~˙21;m;n+˙2a.PEH;1in(4.6)canbelowerboundedby:PEH;1>PLBEH;1=ˆ1M1NXn=1(nld3Ps;njg1;nj2+n~˙21;m;n)+N˙2a:(4.16)From(4.12)and(4.16),theallocatedpowerP1;nislowerboundedby:P1;n>c1ˆ1+1;(4.17)75wherec1="NPn=1nld3Ps;njg1;nj2+n~˙21;m;n+N˙2a#jAj;(4.18)and1=PSU1jAj:(4.19)From(4.17),therateR1in(4.14)canbefurtherboundedby:R1>RLB1=M1MXn2Alog2 (1ˆ2)(c1ˆ1+1)ld4jf1;nj2(1ˆ2)~˙22;m+˙2b!:(4.20)Similarly,wehave:R2>RLB2=M1MXn2Alog2 (1ˆ1)(c2ˆ2+2)ld4jf2;nj2(1ˆ1)~˙21;m+˙2b!;(4.21)wherec2="NPn=1nld2Ps;njg2;nj2+n~˙22;m;n+N˙2a#jAj;(4.22)and2=PSU2jAj:(4.23)764.3.3.2SuboptimalSolutionsInsteadofmaximizingthesum-ratein(4.11),wemaximizethesum-ratelowerbound,whichcanbeformulatedas:maxˆ1;ˆ2RLB1+RLB2s.t.0ˆ11;0ˆ21;(4.24)whichisequivalentto:maxˆ1;ˆ2(1ˆ2)(c1ˆ1+1)(1ˆ2)~˙22;m+˙2b(1ˆ1)(c2ˆ2+2)(1ˆ1)~˙21;m+˙2bs.t.0ˆ11;0ˆ21:(4.25)Theoptimalpowersplittingratiosfor(4.25)canbeobtainedas:ˆ1=24~˙21;m+˙2b˙bq˙2b+(1+1=c1)~˙21;m~˙21;m35+;(4.26)andˆ2=24~˙22;m+˙2b˙bq˙2b+(1+2=c2)~˙22;m~˙22;m35+;(4.27)where[x]+=max(x;0).Thecomputationalcomplexityofthissuboptimalschemeismuchlowerthantheoptimalone,whichrequiresthree-dimensionalexhaustivesearchtogetherwithconvexoptimization.774.4DiscussionsonWorstJammingInterferenceinPUEAInthissection,wediscusstheworst-caseinterferencefortheSUs,whichaimstominimizethemaximumsum-rate.Wewillshowthatforthesecondaryusers,equalpowerinterfer-encefromthemalicioususeristheworstinterferenceforweakjamming,andnearlytheworstinterferenceforstrongjamming,whiletheCSI-assistedinterferenceistheworst-caseinterferenceforstrongjamminginthehighSNRregion.Toseethis,considerthefollowingoptimalinterferencepowerallocationproblemforsum-rateminimizationforSU1:min˙2MU;nR1=Xn2Alog2 1+jHnj2P1;njH2;m;nj2˙2MU;n+˙2!s.t.Xn2A˙2MU;n=˙2MU;(4.28)whereHn=pld4f1;nand˙2=˙2a+˙2b.˙2MU;n=n˙2MU;nandH2;m;ndenotetheallocatedjammingpowerattheMU1,andthefrequencyresponseofthenthsubchannelfromMU1toSU2,respectively.Here,wehaveomittedtheconstantM1Mandassumedthatthereisnoenergyharvesting,i.e.,ˆ2=0.Forweakinterference,wherejH2;m;nj2˙2MU;n˝jHnj2P1;n;8n,andinthehighSNR78region,wehave:R1=Xn2Alog2 1+jHnj2P1;njH2;m;nj2˙2MU;n+˙2!'Xn2Alog2 jHnj2P1;njH2;m;nj2˙2MU;n!=log2Yn2A jHnj2P1;njH2;m;nj2˙2MU;n!(4.29)Hence,minimizingthesum-rateisequivalenttomax˙2MU;nYn2AjH2;m;nj2˙2MU;n(4.30)Itcanbeshownthattheoptimalsolutionis˙2MU;n=˙2MUjAj;8n2A,whichimpliesthatequalpowerinterferenceistheworst-caseinterferenceforweakinterference.Forstronginterference,wherejH2;m;nj2˙2MU;n˛jHnj2P1;n;8n,andinthehighSNRregion,wehave:R1=Xn2Alog2 1+jHnj2P1;njH2;m;nj2˙2MU;n+˙2!'log2(e)Xn2AjHnj2P1;njH2;m;nj2˙2MU;n;(4.31)wherewehaveusedtheapproximationthatlog2(1+x)'xlog2(e)forsmallx.79Theoptimalinterferencepowercanbedeterminedbysolvingthefollowingproblem:min˙2MU;nXn2AjHnj2P1;njH2;m;nj2˙2MU;ns.t.Xn2A˙2MU;n=˙2MU;(4.32)whichisgivenby:˙20MU;n=jHnjpP1;n=jH2;m;njPk2AjHkjpP1;k=jH2;m;kj˙2MU;8n2A:(4.33)Infact,from(4.32),letan=jHnj2P1;njH2;m;nj2.Usingthewellknowninequality:PnxnPnanxnPnpan2,wehave:Xn2A˙2MU;nXn2Aan˙2MU;n Xnpan!2;(4.34)wheretheequalityholdsifandonlyif˙2MU;n=cpan,wherecisaconstanttomeettheconstraintPn2A˙2MU;n=˙2MU.Wecanthenobtaintheresultin(4.33).Theresultaboveshowsthat:understronginterference,theoptimalstrategyforthemalicioususeristoadjustitspowerleveloneachwhitespacesubcarrierbasedontheknowledgeofchannel,namelyCSI-assistedinterference.However,itisnotttoshowthattheperformancegapbetweentheequalpowerinterferenceandCSI-assistedadaptiveinterferenceissmall.Thesum-rateperformancegap80is:R1'Xn2AjHnj2P1;njH2;m;nj2˙2MU;nXn2AjHnj2P1;njH2;m;nj2˙20MU;n=1˙2MUjAjXn2AjHnj2P1;n=jH2;m;nj2(Xn2AjHnjqP1;n=jH2;m;nj)2;(4.35)whichissmallforlarge˙2MUunderstronginterference.Discussions:Fromthederivationsabove,wecanseethatforpracticalsystems,equalpowerallocationatthemalicioususerisnearlyoptimalintermsofsum-ratedegradation.Inotherwords,duetotheabsenceofCSIinformation,theworstjammingfortheSUsiswhenthemalicioususerperformsequalpowerallocationoverallthewhitespacesubcarriers.4.5SimulationResultsInthissection,weevaluatetheperformanceoftheproposedenergyharvestingscheme.Inthesimulation,thepathlossattenuationfactorisgivenbyld=d,whereisthepathlossexponentanddrepresentsthedistancesbetweenthenetworknodes.Moreover,theprimaryuserisassumedtobetransmittingwithequalpowerover112subcarriersfrom128totalsubcarriers.Theremaining16subcarriersarewhitespacesusedforsecondaryusertransmissions.Itisalsoassumedthatthemalicioususerhasperfectknowledgeofthelocationofthewhitespacesubcarriers,andperformsequalpowerallocationoverallthesewhitespacesubcarrierstoachievetheworstinterference.Tomitigatetheinterferencefromthemalicioususer,eachsecondaryuserneedstoincreaseitsowntransmitpowersothatthereceivedsignalpowerattheothersecondaryuserishigherthanthereceivedjammingpower.Forthispurpose,eachsecondaryuserharvestsenergyfromboththePUEAsignals81Table4.1:SystemParametersSamplesizeM100NumberofsubcarriersN128Numberofwhitespace&interferencesubcarriers16DistancebetweenPU1andSCU1(d1)500mDistancebetweenPU1andSU1(d3)50mDistancebetweenPU1andSU2(d2)50mPU1totaltransmitpower(over112subcarriers)2050dBmSU1totaltransmitpower(over16subcarriers)10dBmSU2totaltransmitpower(over16subcarriers)10dBmMU1totaltransmitpower(over16subcarriers)10,20dBmNoisevariancepersubcarrier60dBmPathlossexponent2Energyconversion0:7andtheprimaryusersignals.ThecompletesimulationparametersarelistedinTable4.1.Todemonstratetheenessoftheenergyharvesting,Figure4.2showsachievablesum-rateoftheproposedenergyharvestingschemeandtheconventionalnon-energyhar-vestingOFDMsystemundertPUandMUtransmitpowerlevels.Itcanbeseenthat,withouttheenergyharvestingscheme,thesum-rateislowandindependentofthetransmitpowerlevels.Thisisattributedtothefactthatsecondaryusersarelimitedtotheirinherentpower.Incontrast,withtheproposedscheme,theachievablesum-rateimprovestly,especiallyunderhighPUtransmitpower.Itshouldbenotedthatwhenweconsiderthepowerdistributedtoeachsubcarrier,theMUtransmitpowerlevelsgiveninTable4.1arecomparabletothePUtransmitpowerlevels.ThisthefactthattheMUisemulatingthePUsignal.Moresp,letPsandPmdenotethetotaltransmitpower(indBm)fortheprimaryuserandmalicioususer,respectively.Then,thePUtrans-mitpowerpersubcarrier(indBm)is(Ps10log10(112))dBm,andtheMUtransmitpowerpersubcarrieris(Pm10log10(16))dBm.Thismeans,intermsofindividualsubcarrier,Ps82(a)WhentheMU1transmitpoweris10dBm.(b)WhentheMU1transmitpoweris20dBm.Figure4.2:Theachievablesum-ratecomparisonoftheproposedOFDM-basedenergyhar-vestingschemeandthatofitsnon-energyharvestingcounterpartversustPUandMUtransmitpowerlevels.isequivalenttoPm+10log10(11216).Inthesimulations,thepowersplittingratiosarecalculatedusingboththeoptimalsolu-tion(throughexhaustivesearch)andthesuboptimalsolution.Moresp,Figure4.3showstheoptimalˆ1andˆ2fortheproposedenergyharvestingschemeversusPUtransmitpower.ItcanbeseenthatasthePUtransmitpowerincreases,theoptimalandsubopti-malpowersplittingratiosalsoincrease.Thisindicatesthattherewouldbenobinharvestingenergyforlowreceivedsignalpower.Ontheotherhand,almostallthereceivedsignalisusedforenergyharvestingunderhighPUtransmitpowertoenhancetheuplinktransmission.Itisalsoobservedthattheperformanceofthesuboptimalschemeisfairlyclosetotheoptimalone,butwithmuchlowercomputationalcomplexity.83Figure4.3:TheoptimalandsuboptimalpowersplittingratiosversusPUtransmitpower.4.6SummaryInthischapter,weconsideredenergyharvestingschemeforreliableandtOFDM-basedCRnetworkoperationunderPUEA.First,weproposedanecommunicationschemeforthesecondaryusersunderPUEAbyexploitingtheenergyharvestingtechniques.Optimalpowersplittingwasconsideredforsum-ratemaximization.Astheoptimalsolutionreliesonmulti-dimensionalexhaustivesearch,weproposedanesuboptimalsolutionwithmuchlowercomputationalcomplexity.Itwasobservedthatthesum-rateofthesec-ondaryusernetworkcanbeimprovedsigtlywiththeenergyharvestingtechnique.Second,weevaluatedtheworst-casePUEAinterferenceintermsofminimizingthesum-rateforthesecondaryusers.Weshowedthatforpracticalsystems,asCSIamongtheSUsisnotavailabletothemalicioususer,theworstjammingfortheSUsiswhenthemalicioususerperformsequalpowerallocationoverallthewhitespacesubcarriers.Simulationresultswere84providedtoillustratetheproposedapproaches.Inadditiontothejammingscenario,energyharvestingcanbeappliedinthebenignenvironmentaswelltoresolvetheenergyscarcityintheIoTnetworks.85Chapter5BlockingProbabilityAnalysisforRelay-assistedOFDMANetworksAlongwiththeemerginghighdensityIoTnetworks,relay-assistednetworksareattractingmoreresearchattentioninrecentyearsduetotheirabilitiestoextendthecoverageareaandimprovetheQoS.Blockingprobability(BP)hasbeenusedasaveryimportantmet-ricinevaluatingtheQoSofthenetwork.Inthischapter,takingthespatialrandomnessoftheIoTnetworkintoconsideration,weinvestigateblockingprobabilityinrelay-assistedOFDMAnetworksusingstochasticgeometry.Moresp,wemodeltheinter-cellinterferencefromtheneighboringcellsateachtypicalnode.Then,wederivethecoverageprobabilityinthedownlinktransmissions,includingboththedirectandrelay-basedtrans-missions.Finally,weclassifytheincomingusersintotclassesbasedontheirdataraterequirements,andcalculatetheblockingprobabilityusingthemulti-dimensionallossmodelbasedontheMarkovchains.Weshowthattheblockingprobabilitycanbereducedbyexploitingrelay-assistedtransmissions.5.1IntroductionIndesigningandanalyzingwirelessnetworks,blockingprobabilityhasbeenusedasaveryimportantmetricinevaluatingtheQoSofthenetwork.Blockingprobabilityistheprob-86abilitythatanarrivinguserisdeniedofserviceduetotnetworkresources.Inliterature,blockingprobabilityhasbeeninvestigatedfortheOFDMAnetworksfromentperspectives.Toprovidemorereliableservicesforuserswithlowreceivedsignalstrength,relaystationsareoftendeployedalongwiththebasestationstoextendthecoverageareaandimprovetheQoS.Asaresult,blockingprobabilityhasbeenstudiedinbothsingle-hopandmulti-hopnetworks.Somerepresentativeworkonblockingprobabilityforthesingle-hopOFDMAnetworkscanbefoundin[36{38].In[36],theauthorstriedtomodelthesubcarrier-allocationsystemusingthebatch-arrivalmodelMX=M=c=c.Thismodelisnottlyaccurateduetotheassumptionthatthesubcarriersareallocatedinbulksbutreleasedonebyone.Therefore,in[37],amorerealisticmodelknownasmulti-classErlanglossmodelwasproposed.Motivatedbythefactthatpowershouldbeconsideredinadditiontosubcarriersasasystemresource,amoregeneralmodelwasproposedin[38],whereboththepowerandthesubcarriersareregardedasthesystemresources.Blockingprobabilityformulti-hopOFDMAnetworksisattractingmoreresearchatten-tionalongwiththeemerginghighdensityIoTnetworks.Forexample,in[39],theauthorsproposedasystemmodeltoevaluatetheblockingprobabilityfortherelay-basedOFDMAcellularnetworks.Intheirmodel,eachcellconsistsofabasestation,locatedatthecenterofthecell,andissurroundingbysixrelaystationswithdeterministic(known)locations.How-ever,aspointedoutin[40],resultsbasedonsuchhighlyidealizedmodelsgenerallymaynotbeveryaccurate.Ontheotherhand,itwasalsoshownin[40]thatmodelingthelocationsofthebase/relaystationsstochasticallyaccordingtoaPoissonpointprocessdepictstherealityandcanachievemorereliableandaccurateresultscomparedtoitsidealizedcounterparts.Moreclosely,stochasticgeometryhasbeenwidelyappliedinmodelingwirelessnetworksin87recentyears[41{44].Forinstance,in[41],stochasticgeometrywasusedtoanalyzerelay-aidedtwo-hopnetworks.In[42],itwasusedtoanalyzemulti-hoptransmissioninad-hocnetworks.However,tothebestofourknowledge,noworkhasbeendoneinliteraturetoin-vestigateblockingprobabilityinrelay-assistedOFDMAnetworksusingstochasticgeometry,whichisthemaincontributionofthischapter.InOFDMAnetworks,usersareassignedtnumberofresourceelements(subcar-riers)tomeettheirraterequirements.Therefore,theincominguserscanbeintotclassesbasedontheirsubcarrierrequirements.Averychallengingproblemincalcu-latingtheblockingprobabilityistoobtainthedistributionofthegroupsizeofthesubcarriersthatanarrivinguserneedstoitsdatarate.Inmostoftheexistingwork[36{38],thisdistributionisassumedtobeknown,usuallyassumedtobeuniform.Thismaynotbeaccurate,especiallyforlargegroupsizes,whichactuallyhaveverylowrequestprobability.Inthiswork,wederivethedistributionofuserrequiredsubcarriergroupsizeandthenobtaintheuserblockingprobabilitybasedonthat.Morespally,wemodeltheinter-cellinterferencefromtheneighboringcellsatatypicalnode.Second,wederivethecoverageprobability(i.e.,theprobabilityofsuccessfultransmission)inthedownlinktrans-missions,includingboththedirectandrelay-basedtransmissions.Third,weclassifytheincomingusersintotclassesbasedontheirsubcarrierrequirements.Thesystemunderconsiderationcanbemodeledusingmulti-dimensionalMarkovchains.Finally,wecalculatetheblockingprobabilityusingthemulti-dimensionallossmodel.88Figure5.1:Theproposedrelay-assistednetworkarchitecture.5.2SystemModelInthissection,wepresenttherelay-assistedsystemmodelfortOFDM-basednet-works.5.2.1NetworkModelWeconsiderarelay-assistedOFDM-basednetwork,asshowninFigure5.1.Inthismodel,thenetworkconsistsofbasestations,relaystations,andendusers,suchasIoTdevices.Toaccountforthespatialrandomnessofthenetwork,thelocationsofthebasestationsaremodeledbyahomogeneousPoissonpointprocessB=fXi;i=0;1;2;:::ginR2withintensityB,andthelocationsoftherelaystationsaremodeledasanotherPPPR=fYigwithintensityR,whichisindependentofB.Itisfurtherassumedthateveryusercan89establishaconnectionwithitsgeographicallyclosestbasestation,eitherthroughadirectsingle-hoppath,orthroughamulti-hoppathwithassistancefromtherelaystations.Moresp,basedontheirlocations,theuserscanbedividedintotwogroupsSDandSR,whereSDdenotesthesetofallEUsthatarelocatedatadistancelessthanrBSfromtheirclosestbasestation,andcancommunicatedirectlywiththeBS.Ontheotherhand,SRrepresentsthesetoftheusersthatarelocatedoutsidethatcircleofradiusrBSoftheBS,andonlyconnecttotheBSindirectlythroughtherelaystations,asshowninFigure5.1.AswillbediscussedinSection5.5,theoptimalvalueofrBScanbedeterminedbyminimizingthesystemblockingprobability.5.2.2TransmissionProtocolDesignIdeally,relaysshouldlieonthelinesegmentbetweenthesourceandthedestinationtoavoidthedetourinrouting.Inaddition,assumingthatthesourcenodeandrelaynodeusethesametransmissionpower,thentheirintervalsshouldbeequalsincethethroughputofthepathisbottleneckedbythehopwiththelongestdistance.Motivatedbythisobservation,weproposearoutingprotocolforrelay-assistedtransmission,asshowninFigure5.2.Moresp,assumethatauserEUiislocatedatadistancedBU>rBSfromitsbasestation,Figure5.2:Theproposed2-hoproutingtopology.90i.e,EUi2SR.Typically,EUishouldsenditsdatatotheRSlocatedinthemiddlepointbetweentheEUiandtheBS,andthentheRSdecodestheinformationandforwardsittotheBS.However,sincethelocationsoftheRSsarerandom,itisnotguaranteedthattherewillbeaRSinthehalfwaypointbetweentransmitterandthereceiver.Therefore,intuitively,EUishouldselectsitsnexthoptoaRSthatislocatednearesttotheoptimalmid-point[65].LetrdenotetheminimumdistancebetweentheoptimalRSandanyotherarbitraryRS,thenfromthelawofcosines,rcanbeexpressedas:r=12q2(d2BR+d2RU)d2BU;(5.1)wheredBU,dBRanddRUrepresentthedistancesbetweentheBStotheuser,theBStotheRS,andtheRStotheuser,respectively.Followingtheresultsinstochasticgeometry[40],theprobabilitydensityfunction(PDF)oftherandomvariablercanbeobtainedas:fr(r)=2ˇRreRˇr2:(5.2)ItshouldbeemphasizedthatifdBUrBS,userEUicommunicatesdirectlywiththebasestation,i.e,nohoppingisrequired.Thismeans,ideally,norelaystationsshouldbedeployedinsidethecircularareasoftheBSs.Here,allRSsareassumedtobehalf-duplexusingdecode-and-forwardmechanism,whichmeansthatthe2-hoptransmissionisdonethroughtwottimeslots(i.e.,usingTDDsharingmode).Inthersttimeslot,BS/UEsendsitsinformationtotheRS,thentheRSdecodestheinformationandforwardsittotheUE/BSinthesecondslot.Moreover,acallwillbedroppedifitiswronglydecodedattheRSs.915.2.3ChannelModelWeconsiderapropagationchannelmodel,wherethereceivedpoweratatypicalnodeofatransmittedsignalatadistancedisPr(d)=Pthl(d),wherePtisthetransmitpower,l(d)isthepath-lossattenuationfunctionofdistanced,andhisthechannel(power)gain.Amodelforthepath-lossis:l(d)=0d;(5.3)whereisthepath-lossexponent,0=(4ˇ=v)2,andvistheelectromagneticwavelength.hisexponentiallydistributedrandomvariablewithunitmean,i.e.,Rayleighfadingiscon-sidered.5.3InterferenceModelingandAchievableRateAnal-ysisInthissection,werstanalyzetheinter-cellinterferencefromtheneighboringcellsatatypicalnode.Then,wecalculatethecoverageprobabilityinthedownlinktransmissions.Afterthat,weobtaintheachievabletransmissionrate.5.3.1InterferenceModelInOFDMAnetworks,thetotalavailablespectrumisdividedintoanumberoforthogonalsubcarrierssuchthatnotwousersbelongingtothesamecellsharethesamesetofsubcarriers.Asaresult,theintra-cellinterferenceatatypicalreceivercanbeavoided.However,inter-cellinterferenceduetotheneighboringcellsshouldbeinvestigated.Considerarelay-assistedOFDMsystemwithNsubcarriers,eachofbandwidthf=BN,92whereBisthesystembandwidth.Tasthesetofactivetransmitters(BSsorRSs)occupyingthesamesubcarrierk.FollowingtheindependentthinningpropertyofthePPP[66],TcanbemodeledasahomogeneousPPPwithintensityB,wherepistheactivityfactoroftheBS.Notethatp=1istheworstcasescenariowhereallthetransmittersareactive.Thesignal-to-interference-plus-noiseratio(SINR)experiencedbyatypicalnode(RSorUE)atadistanced0fromatransmitterX02Toversubcarrierkcanbeexpressedas:SINRk=Pt;khkl1(d0)Ik+W;(5.4)whereIk=Xi2TnX0Pt;khk;il1(di);(5.5)istheaggregateinterferencefromallactivetransmittersoversubcarrierk,excepttheservingtransmitterX0.W=fN0isthetotalnoisepoweroversubcarrierksub-band,andN0isthepowerspectraldensityofthenoise.Inaninterferencelimitedscenario,Wissmallcomparedtotheinterferencesfromalltheactivetransmitters,anditcanbeignored.TheLaplacetransform(LT)ofthePDFoftherandomvariableIkevaluatedatscanbeobtainedas[66]:LIk(s)=EfesIkg=exp2ˇB(s0)2ˇsin(2ˇ)g;(5.6)whereitisassumedthatPt;k=1.Hence,withoutlossofgenerality,weconsiderunittransmitpowerthroughoutthischapter.Also,fornotationalsimplicity,weomitthesubscriptkused93toemphasizetheexplicitreferencetoasinglesubcarrier.5.3.2CoverageProbabilityInthissubsection,westudytheprobabilityofsuccessfultransmission(i.e.,thecoverageprobability)indownlinktransmissionincludingboththedirectandrelay-basedtransmis-sions.ThecoverageprobabilityisastheprobabilitythatthereceiveSINRatatypicalnodeisaboveacertainthreshold,whichisthesameasthecomplementarycumulativedistributionfunction(CCDF)oftheSINR,i.e.,Pc()=PrfSINR>g:(5.7)5.3.2.1CoverageProbabilityofDirectTransmissionWithoutlossofgenerality,weassumethatthebasestationunderconsiderationislocatedattheorigin,asshowninFigure5.2.Then,thecoverageprobabilityoftheBS-UElinkcanbeobtainedas:Pc;BU(BUjdBU)=PrfSINRBU>BUg=EInPrfhl1(dBU)I>BUgo=EInPrfh>BUIl(dBU)go=LI(BUl(dBU))=exp2ˇB(BU)2d2BUˇsin(2ˇ)g;(5.8)94whereweusedtheassumptionthathsExp(1),L()istheLaplacetransformevaluatedin(5.6).5.3.2.2CoverageProbabilityof2-hopTransmissionAsmentionedearlier,basedonthedistancebetweentheuseranditsassociatedbasestation,incasewheredBU>rBS,theusermayattempttoconnecttotheBSwiththehelpoftheRS.Therefore,thecoverageprobabilityofthedownlinktransmissionfromthebasestationtotherelaystation(i.e.,BS-RSlink)canbeexpressedas:Pc;BR(BRjdBR)=PrfSINRBR>BRg=exp2ˇB(BR)2d2BRˇsin(2ˇ)g:(5.9)OncethedataisdecodedcorrectlyattheRS,theRSfurtherforwardsittotheintendeduser.ThecoverageprobabilityoftheRS-EUlinkcanbeexpressedas:Pc;RU(RUjdRU)=PrfSINRRU>RUg=exp2ˇB(RU)2d2RUˇsin(2ˇ)g:(5.10)Assumingthetwolinksareindependent,thenthecoverageprobabilityofthe2-hoptransmissioncanbeobtainedas:Pc;BRU(BU;RUjdBR;dRU)=PrfSINRBU>BUgPrfSINRRU>RUg=exp2ˇB(~)2ˇsin(2ˇ)(BU)2d2BR+(RU)2d2RUg:(5.11)95Withoutlossofgenerality,lettherequiredSINRthresholdsforbothlinksbethesame,i.e.,BU=RU=~,thenPc;BRUcanbeto:Pc;BRU(~jdBU;r)=exp2ˇB(~)2ˇsin(2ˇ)(d2BR+d2RU)g=exp2ˇB(~)2ˇsin(2ˇ)(d2BU=2+2r2)g;(5.12)whichcanbeaveragedoverrtoyield:Pc;BRU(~jdBU)=Z10Pc;BRU(~jdBU;r)fr(r)dr=R+Reˇd2BU=4;(5.13)where=4B(~)2ˇsin(2ˇ).Hence,thecoverageprobabilityoftherelay-assistedsystemcanbeexpressedas:Pc=8><>:Pc;BU(BUjdBU);dBUrBS;Pc;BRU(~jdBU);dBU>rBS:(5.14)5.3.3AchievableTransmissionRateHere,weconsiderthesuccessfulachievablerateinthedownlinkrelay-assistednetworks.AccordingtoShannon-Hartleytheorem,givenSINR=,themaximumrateofinforma-tionthatcanbereliablysentoveragivenlinkwithunitbandwidth(i.e.,spectralinbit/s/Hz)islog2(1+).Asaresult,thesuccessfultransmissionpersubcarrieratrate96log2(1+BU)fortheBS-EUlinkcanbeobtainedas:Rb;BU=flog2(1+BU)Pc;BU(BUjdBU)=flog2(1+BU)e2ˇB(BU)2d2BUˇsin(2ˇ):(5.15)Similarly,theachievablerateforthe2-hoptransmissioncanbeobtainedas:Rb;BRU=12flog2(1+~)Pc;BRU(~jdBU)=Rf2(+R)log2(1+~)eˇd2BU=4:(5.16)Notethattheratefortherelay-basedtransmissionishalfthatofthedirecttransmissionbecausethedatarequirestwotimeslotstoreachthedestinationforthe2-hoptransmission.Theachievabletransmissionratefortherelay-assistedsystemcanbeexpressedas:Rb=8><>:Rb;BU;dBUrBS;Rb;BRU;dBU>rBS:(5.17)5.4BlockingProbabilityAnalysisInthissection,weinvestigatetheblockingprobabilityoftherelay-assistedsystem.First,wemodelthetraoftheusersandclassifythemintotclassesbasedontheirrequiredsubcarriers.Weconsiderthequeueingsystemtobemulti-dimensionalwithmultipleclasses.Then,wecalculatetheblockingprobabilityusingthemulti-dimensionallossmodel.975.4.1ResourceAllocationandTModelingInOFDMAnetworks,usersareassignedtnumberofsubcarrierstotheirraterequirements.Infact,evenifallusersrequestthesamerateR,duetotherelativedistancesbetweentheusersandthebasestation,someuserswouldexperiencepoorSINRlevelsandhencewouldrequiremoresubcarrierstosatisfytheirtransmissionrequirements.Toseethis,letRbetherateachievedusingMsubcarriersontheBS-UElink.Then,from(5.15),wehave:R=fMXi=1log2(1+BU;i)Pc;BU(BU;ijdBU):(5.18)AssumethattheSINRtargetsareequalforallMsubcarriers,then,Mcanbeexpressedas:M=Rflog2(1+BU)Pc;BU(BUjdBU)=Rflog2(1+BU)e2ˇB(BU)2d2BUˇsin(2ˇ):(5.19)Itisclearthat,forgivenRandBU,asthedistancebetweenthebasestationandtheuserincreases,Malsoincreases.Thatis,underthesamedataraterequirementR,usersthatareclosetothebasestationneedfewersubcarriers,whileusersthatarefarawayneedmoresubcarrierstocompensateforthelowRSS.Thisexplainswhycell-edgeusersconsumemoreresourcesthanthecell-centerusers.Motivatedbytheobservationabove,wecanclassifytheincomingusersintotclassesbasedontheirsubcarrierrequirements,oralternativelybasedontheirrelativedis-tancestothebasestation,asfollows.First,lettheentiredistancerangebedividedintoLnon-overlappingconsecutiveintervals/classes,denotedbyCj,j=1;2;:::;L.Thus,auseratadistancedBUfromitsservingbasestationintherange[dBU;j1;dBU;j)belongsto98classCj.Werefertothisuserasclass-juser.Second,eachclasswillbeallocatedacertainnumberofsubcarrierstosatisfyitsraterequirement.Withoutlossofgenerality,weassignMjsubcarrierstoclassCj,whereMjisthemeannumberofsubcarriersrequiredtoachievetherateRfordBUintherange[dBU;j1;dBU;j).P(CjjdBUrBS)astheprobabilitythatanarrivinguserbelongstoclassCjandrequiresMjsubcarriersgiventhatitislocatedwithinthebasestationcoverageregion,whichcanbeobtainedas:P(CjjdBUrBS)=P(d0;j1dBU<d0;jjdBUrBS)=P(d0;j1dBU<d0;j;dBUrBS)P(dBUrBS)=Rmin(d0;j;rBS)min(d0;j1;rBS)fdBU(dBU)ddBURrBS0fdBU(dBU)ddBU=eBˇmin(d20;j1;r2BS)eBˇmin(d20;j;r2BS)1eBˇr2BS;(5.20)wherefdBU(dBU)=2ˇBdBUeBˇd2BUisthePDFofthedistancebetweenauseranditsnearestbasestation.d0;j1andd0;j,whichcanbecalculatedfrom(5.19),representthedistancesbetweentheuserandthebasestationtoachievetherateRwithMj1andMj,respectively.P(CjjdBU>rBS)canbecalculatedsimilarly.Usingthelawoftotalprobability,theprobabilitythatanewuserbelongstoclassCj99andrequiresMjsubcarrierscanbeobtainedas:P(Cj)=eBˇmin(d20;j1;r2BS)eBˇmin(d20;j;r2BS)+eBˇmax(d21;j1;r2BS)eBˇmax(d21;j;r2BS):(5.21)5.4.2BlockingProbabilityCalculationWeassumethatusersarrivetothenetworkaccordingtoaPoissonprocesswithmeanarrivalrate,anddepartwithservicerate,inwhichtheservicetimes(holdingtimes)areindependentlyandexponentiallydistributedwithmean1.Asdiscussedearlier,basedontheusersraterequirements,eachuserisallocated,bythebasestation,acertainnumberofsubcarriersfromthetotalNsubcarriers.Thatis,aclass-juserwouldbeassignedMjsubcarriers.IftheavailablenumberofsubcarriersislessthanMj,thenthisuserwillbeblockedfromaccessingthenetworkandmaytrylater.Therefore,ablockingoccurswhenauserisdeniedofserviceduetothetnetworkresources.Sinceeachusercanrequestmultiplesubcarriersandreleasethemsimultaneouslyuponthecompletionoftheservice,thentherelay-assistedsystemcanbemodeledusingmulti-dimensionalMarkovchains.Moresp,letjdenotethearrivalrateoftheusersbelongingtoclassCjandrequestingMjsubcarrierstomeettheirraterequirements.Inthiscase,j=P(Cj),whereP(Cj)obtainedin(5.21).Notethat=Pjjforj=1;2;:::;L.Thestatespace(i.e.,thesetofallallowablestates)oftheexistingusersinthesystemcanbeexpressedas:S=fn:LXj=1njMjNg;(5.22)100wheren=(n1;n2;:::;nL)representsastatewithn1usersfromclassC1,n2usersfromclassC2,andsoon,thattheinequalityin(5.22).Thestationarydistributionofncanbeexpressedinaproductformas[67]:ˇ(n)=1G(N;L)LYj=1ˆnjjnj!;(5.23)whereG(N;L)=Xn2SLYj=1ˆnjjnj!;ˆj=j:(5.24)Now,sincethesystemhasmultipleclassesandeachclassrequiresacertainnumberofsubcarriers,theneachclassexperiencestblockingprobability.LetPB;jdenotetheclassCjblockingprobability,whichistheprobabilitythataclass-juserarrivesandlessthanMjsubcarriersavailableinthesystem.PB;jcanbeobtainedas:PB;j=Xn2Sjˇ(n);(5.25)whereSj=fn:NMj<PLj=1njMjNgdenotesthesetofblockingstatesforclass-juser.TheentiresystemblockingprobabilityPBcanbeobtainedas:PB=1ˆLXj=1ˆjPB;j;(5.26)whereˆ==Pjˆjistheutilizationfactor.Duetothelargestatespaceofthesystem,whichisgenerallyintheorderofO(NL),com-putingtheblockingprobabilityin(5.25)directlyiscomputationallyinet.Therefore,to101reducethecomputationalcomplexity,fornumericalevaluation,weusearecursivealgorithm,knownasKaufman-Robertsalgorithm[68,69],tocalculatetheblockingprobability.Table5.1:SimulationParametersNumberofsubcarriersN256NumberofclassesL200rateR0:2MbpsSubcarrierbandwidthf180KHzBasestationsintensityB1105=m2RelaystationsintensityR1105=m2Basestationactivityfactorp1SINRthresholdsBU=~5dBMeanarrivalrate3Meanservicerate1Pathlossexponent45.5SimulationResultsInthissection,wecarryouttheQoSperformanceanalysisofrelay-assistedsystembasedontheproposedmethodforBPevaluation.Moresp,weobtaintheoptimalvalueofrBS,denotedbyrBS;Opt,thatminimizesthesystemblockingprobabilityPB.Then,usingrBS;Opt,wecalculatetheclassdistributionaswellastheclassblockingprobability.ThesimulationparametersarelistedinTable5.1.Example1:TheoptimalbasestationrangeInthisexample,weobtainrBS;OptthatachievesminimumblockingprobabilityPB,asshowninFigure5.3.ItcanbeseenthatforsmallvaluesofrBS,theblockingprobabilityisrelativelyhighsincemostoftheuserscommunicatewiththebasestationindirectlythroughtherelaystations.AsrBSstartsincreasing,moreuserswillbelocatedinsidethebasestationcoverageregionandestablishadirectconnectionwiththebasestation,whichreducestheblockingprobability.WhenrBS102Figure5.3:Example1:ThesystemblockingprobabilityPBversusrBS.Here,rBS;Opt=225:36m.becomeslarger,PBstartsincreasingagain,sincetheusersthatarefarawaywouldconnecttothebasestationdirectly,buthavelowcoverageprobability.Thus,morenetworkresourceswouldbereleasedforsuchuserstomeettheirservicerequirements.Itcanalsobeobservedthat,whenrBSisverylarge,PBstaysconstant.Inthiscase,allusersarelocatedwithingthebasestationcoveragearea.Fromthediscussionsabove,itisconcludedthatdeployingrelaystationsalongwiththebasestationsextendthecoverageareaandimprovetheQoS,i.e.,reducetheblockingprobability.Example2:ClassdistributionandclassblockingprobabilityUsingrBS;OptobtainedinExample1,weevaluatethedistributionofthenumberofsubcarriersthatanarrivinguserrequirestomeetitsraterequirement.Then,basedonthisdistribution,wecalculatetheclassblockingprobabilityPB;j,asdepictedinFigure5.4.ItcanbenotedfromFigure5.4thatasthegroupsizeofthenumberofsubcarriersincreases,theclass103Figure5.4:Example2:ClassdistributionandclassblockingprobabilityversusMjusingrBS;Opt=225:36m.probabilitydecreases.Thisisincontrastwithmostoftheexistingwork,whichassumesuniformdistribution[36{38].ItcanalsobeseenfromFigure5.4thattheclassblockingprobabilityPB;jisamonotonicallydecreasingfunctionwiththeclasssizes,sinceusersthatarerequiringlargenumberofsubcarriersaremorelikelytobeblockedthanthosethatarerequiringfewersubcarriers.5.6SummaryInthischapter,westudiedblockingprobabilityinrelay-assistedOFDMAnetworksusingstochasticgeometry.First,wemodeledtheinter-cellinterferencefromtheneighboringcellsatatypicalnode.Second,wederivedthecoverageprobabilityinthedownlinktransmissions,includingboththedirectandrelay-basedtransmissions.Third,wetheincoming104usersintotclassesbasedontheirsubcarrierrequirements,andcalculatedthesys-temblockingprobability.Itwasshownthat,withtherelay-assistedsystem,theblockingprobabilitycanbereducedandthereforethenetworkperformancecanbeimproved.105Chapter6ConclusionsandFutureWork6.1ConclusionsIoTisthenextimpetusinthedevelopmentofnewwirelessnetworks.Themainchallengesherelieinespectrumsharing,highnetworkdensity,andthescarceenergybudgetandthelowcostrequirementsofIoTapplications.Wetriedtoaddressthesechallengesfromthreeaspects:(1)Secureandespec-trumsharinginthelowerfrequencybands;(2)EnergyttransmissionforIoTdevicesthroughenergyharvesting;and(3)QoSevaluationinthehighdensityIoTnetworks.Themainconclusionsaresummarizedinthefollowing.First,wedevelopedasecureandtspectrumsharingschemeintheHDTVbands.Westartedwithsingle-carriertransmission,wherethesignalisprocessedintheentirespec-trumandhenceallowsfull-banddetection.InChapter2,weconsideredfull-bandPUEAdetectionandproposedareliableAES-assistedDTVscheme,whereanAES-encryptedref-erencesignalisgeneratedattheDTVtransmitterandusedasthesyncbitsoftheDTVdataframes.Byallowingasharedsecretbetweenthetransmitterandthereceiver,therefer-encesignalcanberegeneratedatthereceiverandusedtoachievefull-bandidenofauthorizedPUs.Moreover,whencombinedwiththeanalysisontheauto-correlationofthereceivedsignal,thepresenceoftheMUcanbedetectedaccuratelynomattertheprimaryuserispresentornot.ItwasshownthatwiththeAES-assistedDTVscheme,boththePU106andMUcanbedetectedwithhighaccuracyandlowfalsealarmrateunderPUEA.Wefurtherdevelopedasecureandtspectrumsharingschemeinmulti-carrierwirelessnetworks.InChapter3,weconsideredsub-bandPUEAdetectionintheOFDM-basedCRnetworks.WeproposedantAES-basedDTVscheme,wheretheexistingreferencesequenceusedtogeneratethepilotsymbolsintheDVB-T2framesisencryptedusingtheAESalgorithmforaccuratesub-bandPUandMUdetection.Withtheproposedscheme,wecandetectPUEAaccuratelyoverallsubcarriersorsub-bandswheretheP2symbolspresent.NumericalresultsshowedthatwiththeAES-basedDTVscheme,boththePUandMUcanbedetectedaccuratelyunderPUEA.Potentially,theproposedschemecanbeapplieddirectlytofutureIoTnetworksformoretspectrumsharingwiththeexistinglowerbandsystems.Then,inChapter4,weproposedanecommunicationschemefortheSUsunderPUEAbyexploitingtheenergyharvestingtechniques.Optimalpowersplittingwasconsideredforsum-ratemaximization.Itwasobservedthatthesum-rateoftheSUnetworkcanbeimprovedtlywiththeenergyharvestingtechnique.Then,weevaluatedtheworst-casePUEAinterferenceintermsofminimizingthesum-ratefortheSUs.Weshowedthatforpracticalsystems,asCSIamongtheSUsisnotavailabletothemalicioususer,theworstjammingfortheSUsiswhentheMUperformsequalpowerallocationoverallthewhitespacesubcarriers.Simulationresultswereprovidedtoillustratetheproposedapproaches.Inadditiontothejammingscenario,energyharvestingcanbeappliedinthebenignenvironmentaswelltoresolvetheenergyscarcityintheIoTnetworks.Finally,inChapter5,takingthespatialrandomnessoftheIoTnetworkintoconsidera-tion,weinvestigatedcoverageprobabilityandblockingprobabilityinrelay-assistedOFDMAnetworksusingstochasticgeometry.Moresp,wemodeledtheinter-cellinterference107fromtheneighboringcellsateachtypicalnode,andderivedthecoverageprobabilityinthedownlinktransmissions,includingboththedirectandrelay-basedtransmissions.Basedontheirdataraterequirements,wetheincomingusersintotclasses,andcal-culatedtheblockingprobabilityusingthemulti-dimensionallossmodel.Itwasshownthattheblockingprobabilitycanbereducedbyexploitingrelay-assistedtransmissions.6.2FutureWorkWeproposethefollowingdirectionsforfutureresearch.SecurityenhancementforIoTnodesundermaliciousattacksInChapters2and3,weproposedsecureandtspectrumsharingschemesthatcanenableIoTnodestoaccuratelyidentifythemaliciousattacks.Foretransmission,however,theycanadopttmethodologiessuchas:Avoidtransmissiononthewhitespacesjammedbymaliciousnodes.Forexample,con-siderthecasewheretheIoTnodesareOFDM-basedtransceivers,thentheycanshapetheirtransmittedsignalthroughproperprecodingdesigntoavoidcommunicationoverthejammedsubcarriers.Weplantocarryoutmoreresearchonthisbyexploitingsecuresymbol-levelcoding,whichcanprovidemoredesignyunderhostilejamming.Fortime-varyingattacks,theprecodershouldbeadaptedaccordinglyfortransmission.Thisnecessitatesthatjammingdetectionneedstobeperformedinreal-time,whichcangenerallybeachievedbyevaluatingthetime-varyingpowerspectrumofthejammingsignal.Exploittechniquesthatareinherentlyjamming-resistant,suchascodedivisionmultipleaccess(CDMA)andfrequencyhopping(FH)techniques.BothCDMAandFHwere108initiallydevelopedforsecuremilitarycommunications.CDMAisparticularlytundernarrow-bandjamming[70],evenifthemalicioususerhopsfrombandtoband.FHbasedsystemsaregenerallyrobustunderwide-bandjamming.Furtherresearchcanbeconductedtodevelopasecuresubcarrierallocationschemeforsecureandcollision-freetransmissionbetweenthesecondaryusers.Whenthemaliciousjammingpatternistime-varying,i.e.,themalicioususerswitchesbetweenwide-bandandnarrow-bandjamming,thetransmitterthenneedstobeadjustedtocombatthecognitivehostileattacks.ReliableIoTnetworkdesignandperformanceevaluationunderhostileenviron-mentsWewillconsidertodevelopnewtechniquestodetecterroneoustransmissionsinmulti-hopnetworks,andexaminewhetheritisanodefailureorlinkfailure.Wewillevaluatetheimpactsofdtnodesonthenetwork,andaimtoperformquantitativeanalysisonnetworkthroughputandreliability.Wewillalsoconsiderthedevelopmentofmoreeroutingprotocoldesigntoenhancenetworkperformanceandreliability.QoSperformanceanalysisforCRnetworksunderPUEAInChapter5,takingthespatialrandomnessoftheIoTnetworkintoconsideration,weinvestigatedblockingprobabilityinrelay-assistedOFDMAnetworksusingstochasticgeom-etry.Itwasshownthattheblockingprobabilitycanbereducedbyexploitingrelay-assistedtransmissions.Motivatedbythisobservation,weplantostudyblockingprobabilityinCRnetworksunderPUEAforbothsingle-hopandmulti-hoptransmissionschemes.109APPENDIX110AppendixProofofParametersusedinSection3.3.3Inthisappendix,wecalculate00,˙00,01,˙01,10,˙10,11,˙11,andtheoptimalvaluesforTk;0andTk;1.UnderH00,boththeprimaryuserandmalicioususerareabsent,resultinginrk;i=nk;i.Itfollowsthat:00=1ME8<:MXi=1nk;ink;i9=;=˙2n;(.1)and˙200canbeobtainedas:˙200=Enj^Rrr;kj2oj00j2=1MEfj~nkj4g(˙2n)2;(.2)whereweassumethatEfjnk;ij4g=Efj~nj4g8i.Similarly,underH01,thereceivedsignalis111representedasrk;i=mk;i+nk;i,andthemean01canbeobtainedasfollows:01=1ME8<:MXi=1(mk;i+nk;i)(mk;i+nk;i)9=;=˙2m;k+˙2n:(.3)Thevariance˙201canbeobtainedas:˙201=Enj^Rrr;kj2oj01j2=1MEfj~mkj4g+Efj~nkj4g+Ef2Rf(~mk)2(~nk)2gg+2˙2m;k˙2n(˙2m;k)2(˙2n)2;(.4)whereweassumethatEfjmk;ij4g=Efj~mj4g,Efjnk;ij4g=Efj~nj4g,Ef2Rf(mk;i)2(nk;i)2gg=Ef2Rf(~m)2(~n)2gg,8i.UnderH10,thereceivedsignalisexpressedasrk;i=pld1Ps;kjhkjsk;i+nk;i,andthemean10canbeobtainedasfollows:10=1ME8<:MXi=1(~Cksk;i+nk;i(~Cksk;i+nk;i)9=;=~C2k+˙2n;(.5)112where~Ck=pld1Ps;kjhkj.Thevariance˙210canbeobtainedas:˙210=Enj^Rrr;kj2oj10j2=1M~C4kEfj~skj4g+Efj~nkj4g+Ef2~C2kRf(~sk)2(~nk)2gg+2~C2k˙2n~C4k(˙2n)2:(.6)Similarly,underH11,thereceivedsignalisrepresentedasrk;i=pld1Ps;kjhkjsk;i+mk;i+nk;i,andthemean11canbeobtainedasfollows:11=~C2k+˙2m;k+˙2n;(.7)and˙211canbeobtainedas:˙211=Enj^Rrrj2oj11j2=1M~C4kEfj~skj4g+Efj~mkj4g+Efj~nkj4g+2~C2kEfRf(~sk)2(~mk)2gg+2~C2kEfRf(~sk)2(~nk)2gg+2EfRf(~mk)2(~nk)2gg+2~C2k˙2m;k+2~C2k˙2n+2˙2m;k˙2n~C4k(˙2m;k)2(˙2n)2:(.8)Similartothethresholdforprimaryuserdetection,followingtheminimumdistancerule,wechooseTk;0=(˙2m;k+2˙2n)=2andTk;1=(2ld1Ps;kjhkj2+˙2m;k+2˙2n)=2asthethresholdsformalicioususerdetection.113BIBLIOGRAPHY114BIBLIOGRAPHY[1]FederalCommunicationsCommission,\Spectrumpolicytaskforcereport,"ETDocketNo.02-135,November2002.[2]B.WangandK.J.R.Liu,\Advancesincognitiveradionetworks:Asurvey,"IEEEJournalofSelectedTopicsinSignalProcessing,vol.5,no.1,pp.5{23,Feb2011.[3]S.Haykin,\Cognitiveradio:brain-empoweredwirelesscommunications,"IEEEJournalonSelectedAreasinCommunications,vol.23,no.2,pp.201{220,2005.[4]I.F.Akyildiz,W.-Y.Lee,M.C.Vuran,andS.Mohanty,\NeXtgeneration/dynamicspectrumaccess/cognitiveradiowirelessnetworks:Asurvey,"ComputerNetworks,vol.50,no.13,pp.2127{2159,2006.[5]M.Thanu,\Detectionofprimaryuseremulationattacksincognitiveradionetworks,"inInternationalConferenceonCollaborationTechnologiesandSystems(CTS),May2012,pp.605{608.[6]FederalCommunicationsCommission,\UnlicensedoperationintheTVbroadcastbandsandadditionalspectrumforunlicenseddevicesbelow900MHzandinthe3GHzband,"ETDocketNo.04-186and02-380,September2010.[7]R.ChenandJ.-M.Park,\Ensuringtrustworthyspectrumsensingincognitiveradionetworks,"inIEEEWorkshoponNetworkingTechnologiesforSoftwaredRadioNetworks,Sept.2006,pp.110{119.[8]T.ClancyandN.Goergen,\Securityincognitiveradionetworks:Threatsandmit-igation,"inCognitiveRadioOrientedWirelessNetworksandCommunications,2008.CrownCom2008.Conferenceon3rdInternational,May2008,pp.1{8.[9]A.Fragkiadakis,E.Tragos,andI.Askoxylakis,\Asurveyonsecuritythreatsanddetectiontechniquesincognitiveradionetworks,"IEEECommunicationsSurveys&Tutorials,vol.15,no.1,pp.428{445,2013.[10]S.Anand,Z.Jin,andK.P.Subbalakshmi,\Ananalyticalmodelforprimaryuseremulationattacksincognitiveradionetworks,"inNewFrontiersinDynamicSpectrumAccessNetworks,2008.DySPAN2008.3rdIEEEonSymposium,2008,pp.1{6.115[11]N.Nguyen,R.Zheng,andZ.Han,\OnidentifyingprimaryuseremulationattacksincognitiveradiosystemsusingnonparametricBayesianIEEETransac-tionsonSignalProcessing,vol.60,no.3,pp.1432{1445,2012.[12]R.Chen,J.-M.Park,andJ.Reed,\Defenseagainstprimaryuseremulationattacksincognitiveradionetworks,"IEEEJournalonSelectedAreasinCommunications,vol.26,no.1,pp.25{37,Jan.2008.[13]Z.Yuan,D.Niyato,H.Li,andZ.Han,\Defenseagainstprimaryuseremulationattacksusingbeliefpropagationoflocationinformationincognitiveradionetworks,"inIEEEWirelessCommunicationsandNetworkingConference(WCNC),2011,pp.599{604.[14]Z.Yuan,D.Niyato,H.Li,J.B.Song,andZ.Han,\Defeatingprimaryuseremulationattacksusingbeliefpropagationincognitiveradionetworks,"IEEEJournalonSelectedAreasinCommunications,vol.30,no.10,pp.1850{1860,2012.[15]Z.Jin,S.Anand,andK.P.Subbalakshmi,\Detectingprimaryuseremulationattacksindynamicspectrumaccessnetworks,"inIEEEInternationalConferenceonCommu-nications,June2009,pp.1{5.[16]||,\Mitigatingprimaryuseremulationattacksindynamicspectrumaccessnetworksusinghypothesistesting,"SIGMOBILEMob.Comput.Commun.Rev.,vol.13,no.2,pp.74{85,2009.[17]C.Chen,H.Cheng,andY.-D.Yao,\Cooperativespectrumsensingincognitiveradionetworksinthepresenceoftheprimaryuseremulationattack,"IEEETransactionsonWirelessCommunications,vol.10,no.7,pp.2135{2141,2011.[18]M.Dang,Z.Zhao,andH.Zhang,\Detectionofprimaryuseremulationattacksbasedoncompressivesensingincognitiveradionetworks,"inInternationalConferenceonWirelessCommunications&SignalProcessing(WCSP),Oct2013,pp.1{5.[19]S.Chen,K.Zeng,andP.Mohapatra,\Hearingisbelieving:Detectingmobileprimaryuseremulationattackinwhitespace,"inProceedingsIEEEINFOCOM,April2011,pp.36{40.[20]S.Jana,K.Zeng,W.Cheng,andP.Mohapatra,\Trustedcollaborativespectrumsensingformobilecognitiveradionetworks,"IEEETransactionsonInformationForensicsandSecurity,vol.8,no.9,pp.1497{1507,Sept2013.116[21]S.L.Ho,J.Wang,W.N.Fu,andM.Sun,\Acomparativestudybetweennovelwitricityandtraditionalinductivemagneticcouplinginwirelesscharging,"IEEETransactionsonMagnetics,vol.47,no.5,pp.1522{1525,May2011.[22]X.Lu,P.Wang,D.Niyato,D.I.Kim,andZ.Han,\Wirelesschargingtechnologies:Fundamentals,standards,andnetworkapplications,"IEEECommunicationsSurveysTutorials,vol.18,no.2,pp.1413{1452,Secondquarter2016.[23]A.Kurs,A.Karalis,R.J.D.Joannopoulos,P.Fisher,andM.Soljacic,\Wirelesspowertransferviastronglycoupledmagneticresonances,"Science,vol.317,no.5834,pp.83{86,2007.[Online].Available:http://science.sciencemag.org/content/317/5834/83[24]L.Varshney,\Transportinginformationandenergysimultaneously,"inIEEEInterna-tionalSymposiumonInformationTheory,July2008,pp.1612{1616.[25]R.ZhangandC.K.Ho,\MIMObroadcastingforsimultaneouswirelessinformationandpowertransfer,"IEEETransactionsonWirelessCommunications,vol.12,no.5,pp.1989{2001,May2013.[26]X.Zhou,R.Zhang,andC.K.Ho,\Wirelessinformationandpowertransfer:Architec-turedesignandrate-energyIEEETransactionsonCommunications,vol.61,no.11,pp.4754{4767,November2013.[27]C.MikekaandH.Arai,Designissuesinradiofrequencyenergyharvestingsystem.INTECHOpenAccessPublisher,2011.[28]X.Lu,P.Wang,D.Niyato,D.I.Kim,andZ.Han,\WirelessnetworkswithRFenergyharvesting:Acontemporarysurvey,"IEEECommunicationsSurveysTutorials,vol.17,no.2,pp.757{789,Secondquarter2015.[29]D.Ng,E.Lo,andR.Schober,\EnergytresourceallocationinmultiuserOFDMsystemswithwirelessinformationandpowertransfer,"inIEEEWirelessCommunica-tionsandNetworkingConference(WCNC),April2013,pp.3823{3828.[30]D.Li,C.Shen,andZ.Qiu,\Sumratemaximizationandenergyharvestingfortwo-wayafrelaysystemswithimperfectCSI,"inIEEEInternationalConferenceonAcoustics,SpeechandSignalProcessing(ICASSP),May2013,pp.4958{4962.[31]S.Lee,R.Zhang,andK.Huang,\Opportunisticwirelessenergyharvestingincognitiveradionetworks,"IEEETransactionsonWirelessCommunications,vol.12,no.9,pp.4788{4799,September2013.117[32]D.T.Hoang,D.Niyato,P.Wang,andD.I.Kim,\OpportunisticchannelaccessandRFenergyharvestingincognitiveradionetworks,"IEEEJournalonSelectedAreasinCommunications,vol.32,no.11,pp.2039{2052,November2014.[33]A.Sultan,\Sensingandtransmitenergyoptimizationforanenergyharvestingcognitiveradio,"IEEEWirelessCommunicationsLetters,vol.1,no.5,pp.500{503,2012.[34]A.andA.Sultan,\Optimalrandomaccessforacognitiveradioterminalwithenergyharvestingcapability,"IEEECommunicationsLetters,vol.17,no.6,pp.1128{1131,June2013.[35]S.Park,H.Kim,andD.Hong,\Cognitiveradionetworkswithenergyharvesting,"IEEETransactionsonWirelessCommunications,vol.12,no.3,pp.1386{1397,March2013.[36]J.-C.ChenandW.-S.E.Chen,\CallblockingprobabilityandbandwidthutilizationofOFDMsubcarrierallocationinnext-generationwirelessnetworks,"IEEECommunica-tionsLetters,vol.10,no.2,pp.82{84,Feb2006.[37]V.Pla,J.Martinez-Bauset,andV.Casares-Giner,\Commentson"callblockingprob-abilityandbandwidthutilizationofOFDMsubcarrierallocationinnext-generationwirelessnetworks","IEEECommunicationsLetters,vol.12,no.5,pp.349{349,May2008.[38]C.PaikandY.S.Suh,\Generalizedqueueingmodelforcallblockingprobabilityandre-sourceutilizationinOFDMwirelessnetworks,"IEEECommunicationsLetters,vol.15,no.7,pp.767{769,July2011.[39]M.Mehta,R.B.Jain,andA.Karandikar,\Analysisofblockingprobabilityinarelay-basedcellularOFDMAnetwork,"WirelessPersonalCommunications,vol.84,no.4,pp.2467{2492,2015.[40]J.G.Andrews,F.Baccelli,andR.K.Ganti,\Atractableapproachtocoverageandrateincellularnetworks,"IEEETransactionsonCommunications,vol.59,no.11,pp.3122{3134,November2011.[41]W.LuandM.D.Renzo,\Stochasticgeometrymodelingandsystem-levelanalysis&optimizationofrelay-aideddownlinkcellularnetworks,"IEEETransactionsonCom-munications,vol.63,no.11,pp.4063{4085,Nov2015.118[42]P.H.J.Nardelli,P.Cardieri,andM.Latva-aho,ofwirelessnetworksunderthoppingstrategies,"IEEETransactionsonWirelessCommunications,vol.11,no.1,pp.15{20,January2012.[43]H.ElSawy,E.Hossain,andM.Haenggi,\Stochasticgeometryformodeling,analysis,anddesignofmulti-tierandcognitivecellularwirelessnetworks:Asurvey,"IEEECommunicationsSurveysTutorials,vol.15,no.3,pp.996{1019,Third2013.[44]F.Baccelli,B.Blaszczyszyn,andP.Muhlethaler,\Analohaprotocolformultihopmobilewirelessnetworks,"IEEETransactionsonInformationTheory,vol.52,no.2,pp.421{436,Feb2006.[45]AT32UC3A3256S.[Online].Available:http://www.atmel.com/devices/at32uc3a3256s.aspx[46]A.Hodjat,D.D.Hwang,B.Lai,K.Tiri,andI.Verbauwhede,\A3.84Gbits/sAEScryptocoprocessorwithmodesofoperationina0.18-CMOStechnology,"inPro-ceedingsofthe15thACMGreatLakessymposiumonVLSI.NewYork,NY,USA:ACM,2005,pp.60{63.[47]S.-Y.LinandC.-T.Huang,\Ahigh-throughputlow-powerAEScipherfornetworkapplications,"inDesignAutomationConference,,2007,pp.595{600.[48]J.AddaandM.Ottaviani,\Digitaltelevision1:Thetransitiontodigitaltelevision*,"September2004.[49]AdvancedTelevsionSystemsCommittee,\A/53:ATSCdigitaltelevisionstandard,part2,"Tech.Rep.,Dec.2011.[50]V.-H.Pham,J.-Y.Chouinard,A.Semmar,X.Wang,andY.Wu,\EnhancedATSCDTVchannelestimation,"inCanadianConferenceonElectricalandComputerEngi-neering,May2009,pp.772{776.[51]W.Stallings,CryptographyandNetworkSecurity:PrinciplesandPractice,5thed.PrenticeHall,Jan.2010.[52]C.MathurandK.P.Subbalakshmi,\DigitalsignaturesforcentralizedDSAnetworks,"inConsumerCommunicationsandNetworkingConference,2007.CCNC2007.4thIEEE,2007,pp.1037{1041.119[53]K.Borle,B.Chen,andW.Du,\Aphysicallayerauthenticationschemeforcounteringprimaryuseremulationattack,"in2013IEEEInternationalConferenceonAcoustics,SpeechandSignalProcessing(ICASSP),2013,pp.2935{2939.[54]W.Burr,\Selectingtheadvancedencryptionstandard,"IEEESecurityPrivacy,vol.1,no.2,pp.43{52,Mar2003.[55]T.GoodandM.Benaissa,\AESasstreamcipheronasmallFPGA,"inCircuitsandSystems,2006.ISCAS2006.Proceedings.2006IEEEonInternationalSymposium,2006,pp.4pp.{.[56]N.SinghalandJ.Raina,\ComparativeanalysisofAESandRC4algorithmsforbetterutilization,"inInternationalJournalofComputerTrendsandTechnology,Aug.2011.[57]D.DasandS.Das,\Primaryuseremulationattackincognitiveradionetworks:Asurvey,"IRACST{InternationalJounralofComputerNetworksandWirelessCommu-nicationsISSN,pp.2250{3501,2013.[58]C.yenOng,J.Song,C.Pan,andY.Li,\Technologyandstandardsofdigitaltele-visionterrestrialmultimediabroadcasting[topicsinwirelesscommunications],"IEEECommunicationsMagazine,vol.48,no.5,pp.119{127,May2010.[59]DigitalVideoBroadcasting,\Framestructurechannelcodingandmodulationforasecondgenerationdigitalterrestrialtelevisionbroadcastingsystem(DVB-T2),"ETSIStd.EN302755V1.3.1,April2012.[60]T.Jiang,T.Li,andJ.Ren,\Towardsecurecognitivecommunicationsinwirelessnet-works,"IEEEWirelessCommunications,vol.19,no.4,pp.82{88,August2012.[61]N.Haller,\TheS/KEYone-timepasswordsystem,"inProceedingsoftheInternetSocietySymposiumonNetworkandDistributedSystemSecurity,February1994.[62]P.S.Mann,IntroductoryStatistics,7thed.Wiley,February2010.[63]X.Zhou,R.Zhang,andC.K.Ho,\Wirelessinformationandpowertransfer:Archi-tecturedesignandrate-energyinIEEEGlobalCommunicationsConference(GLOBECOM),Dec2012,pp.3982{3987.[64]Z.-Q.LuoandW.Yu,\Anintroductiontoconvexoptimizationforcommunicationsandsignalprocessing,"IEEEJournalonSelectedAreasinCommunications,vol.24,no.8,pp.1426{1438,Aug2006.120[65]Z.Lin,Y.Li,S.Wen,Y.Gao,X.Zhang,andD.Yang,\Stochasticgeometryanalysisofachievabletransmissioncapacityforrelay-assisteddevice-to-devicenetworks,"in2014IEEEInternationalConferenceonCommunications(ICC),June2014,pp.2251{2256.[66]F.BaccelliandB.Blaszczyszyn,StochasticGeometryandWirelessNetworks:VolumeITheory(FoundationsandTrends(r)inNetworking).NowPublishersInc,2009.[67]R.R.Mazumdar,\Performancemodeling,lossnetworks,andstatisticalmultiplexing,"SynthesisLecturesonCommunicationNetworks,vol.2,no.1,pp.1{151,2009.[68]J.Kaufman,\Blockinginasharedresourceenvironment,"IEEETransactionsonCom-munications,vol.29,no.10,pp.1474{1481,Oct1981.[69]J.W.Roberts,\Aservicesystemwithheterogeneoususerrequirements,"inPerfor-manceofDataCommunicationssystemsandtheirapplications,vol.29,no.10.Ams-terdam,TheNetherlands:North-Holland,1981,pp.423{431.[70]T.Li,Q.Ling,andJ.Ren,\Physicallayerbuilt-insecurityanalysisandenhancementalgorithmsforCDMAsystems,"EURASIPJournalonWirelessCommunicationsandNetworking,vol.2007,no.1,p.083589,Jul2007.[Online].Available:http://jwcn.eurasipjournals.com/content/2007/1/083589121