ToappearinISOC-SNDSS' 
`AnEmpiricalStudyofSecureMPEGVideoTransmissions
`IskenderAgiandLiGong
`SRIInternational
`ComputerScienceLaboratory
` RavenswoodAvenue
`MenloPark,California ,U.S.A.
`e-mail:iskender,gong@csl.sri.com
`URL:http://www.csl.sri.com
`Abstract
`MPEG(MovingPicturesExpertGroup)isanin-
`dustrialstandardforvideoprocessingandiswidely
`usedinmultimediaapplicationsintheInternet.How-
`ever,nosecurityprovisionisspeci(cid:12)edinthestandard.
`Weconductedanexperimentalstudyofpreviouslypro-
`posedselectiveencryptionschemesforMPEGvideo
`security.Thisstudyshowedthatthesemethodsare
`inadequateforsensitiveapplications.Wediscussthe
`tradeo(cid:11)sbetweenlevelsofsecurityandcomputational
`andcompressione(cid:14)ciency.
` IntroductionandMotivation
`MPEG(MovingPicturesExpertGroup)isanin-
`dustrialstandardforvideoprocessing.Asnetwork
`bandwidthbecomesmoreabundant,MPEGandits
`variantvideocodingtechniquesareincreasinglyused
`inapplicationsovertheInternet,suchasvideodissem-
`inationanddesktopconferencing.Awarenessofsecu-
`ritythreatsandincidentshavepromptedresearchers
`toexaminesecurewaysofencryptingMPEGtra(cid:14)c.
`NoprovisionforencryptionwasincludedforMPEG
`transmissions.TheverynatureofMPEGencoding|
`theencodingofinherentlyspatiallyandtemporally
`
`

`

`ISOC-SNDSS' 
`AnEmpiricalStudyofSecureMPEGVideoTransmissions
`cludingsatellite,coaxialcableorevenCD-ROM.An
`gardtopreviousorfutureframes.Theencodertakes
`MPEGdecoderreceivesthisbitstreamanddecodes
`theDCTofeach(cid:2)block,transformingthatblock
`ittoreproduceafacsimileoftheoriginalvideose-
`intoitsfrequency-basedrepresentation.Sincespa-
`quence.WediscussMPEG- encoding,butbasedon
`tialsignalshavehighlocalautocorrelation,mostof
`ourexperimentstheseresultsalsoapplytoMPEG-.
`thefrequencyinformationislocatedinthelowerfre-
`quencyterms.Theencodereliminatesmanyofthe
`MPEG- videosequencestypicallyuseaframe
`higher-ordertermsusingquantization.Theuseof
`sizeof (cid:2)pixelsdisplayedatarateof-
`lossycompressionincreasesthecompressiontremen-
` framespersecond(FPS).Onceencoded,thebit-
`dously.TheI-frameencodingusesthesamescheme
`streamofthistypicalvideosequenceisdesignedto
`asJPEG(JointPicturesExpertGroup)encodingof
`reproduceaVHS-qualityvideoatatransmissionrate
`stillframes.ThisencodedI-frameisusedasthe
`of .Mb/sec.TheMPEGencoderisnotlimitedto
`motion-estimationreferencefortheotherforward-and
`thesetypicalvaluesandcanencodelargerorsmaller
`bidirectional-predictedframes.
`imagesizesandatvaryingqualitysettings,whichaf-
`fectthetransmissionrateaccordingly.
`Forward-predicted(P)framesareencodedwithref-
`erencetothemostrecentpreviousI-orP-frame.
`MPEGencodingofavideosequencerequiresmany
`Eachmacroblockundergoesmotion-estimationsearch
`compressionsteps.Eachframeinthevideosequence
`to(cid:12)ndtheso-calledbest-(cid:12)tvectorrelativetotherefer-
`iscompressedusingacombinationofblock-basedmo-
`enceframe.Thisvectoristhentransmittedalongwith
`tioncompensationtotakeadvantageofinterframe
`theerrorbetweenthemacroblockandthemotion-
`temporalredundancy,anddiscretecosinetransform
`compensatedmacroblockfromthepreviousframeen-
`(DCT)-basedcompressiontotakeadvantageofin-
`codedusingtheDCT.Theuseofexhaustivesearch
`traframespatialredundancy[].Eachframe(orpic-
`algorithmsformotionestimationmakesMPEGen-
`ture)inthesequenceisbrokeninto(cid:2)pixelblocks
`codingmuchmorecomputationallydemandingthan
`forinterframeDCTcompressionand (cid:2) pixel
`decoding.Macroblockswithnomotionandnoerror
`macroblocksforintraframemotioncompensation.A
`aremarkedasskippedblocks,andnofurthercodingis
` (cid:2) macroblockalsoincludestwo(cid:2)chromi-
`performed.
`nanceblocksinadditiontothefourluminanceblocks
`foratotalofsixblockspermacroblock.
`Bidirectional-predicted(B)framesarecodedwith
`Anumberoftheseblocksstrungtogetherforma
`referencetoboththeimmediatelypreviousIorPref-
`slice(ofaframe)thatactsasaresynchronizationunit.
`erenceframeandtheimmediatelyfutureIorPrefer-
`Forstandardsizeframesthenumberofslicesperframe
`enceframe.Themotion-estimationandencodingpro-
`isone.Largerframesizesmayusemoreslicesper
`cedureforB-framesissimilartotheprocedureused
`frame.Thentheseslicescombinetoformapicture.
`toencodetheP-frames:itincludesthecalculationof
`thebest-(cid:12)tmotionvectorandthecodingoftheerror
`Anumberofthesepicturesaregroupedtogether
`termsusingtheDCT.
`toformarandomaccessunitsothatthevideocan
`beviewedeitherforwardorbackward.Thisgroup
`ThesequenceofI-,P-andB-framesde(cid:12)neswhatis
`ofpictures(GOP)haslittleornodependenceon
`calledtheframepatternofaGOP.TheGOPalways
`thepicturesinadjacentGOPs.AGOPconsistsof
`beginswithanI-frame.Figureshowstheframepat-
`threetypesofframes:intracodedframes(I),motion-
`ternofatypicalMPEGsequence.Theframepattern
`estimatedforwardpredictedframes(P),andmotion-
`lengthisthenumberofP-andB-framesuntilthenext
`estimatedbidirectionalpredictedframes(B)whose
`I-frame.AGOPthatendswithaP-framehasnoen-
`framedependencyisshowninFigure .
`codingdependenceontheframesinthenextGOP.
`However,ifaGOPendsinaB-frame,thentheencod-
`ingofthatB-framerequiresaccesstothe(cid:12)rstframe
`ofthenextGOPinthesequence.
`Figure :MPEGframedependency
`Figure:MPEGframepattern
`I-framesareencodedblockbyblock,withoutre-
`Copyright(c) IEEE.
`
`
`
`I
`
`B
`
`P
`
`GOP
`
`I
`
`B
`
`B
`
`P
`
`B
`
`B
`
`P
`
`GOP
`
`Group Of Pictures (GOP)
`
`DivX, LLC Exhibit 2017
`Page 2017 - 2
`Netflix Inc. et al. v. DivX, LLC, IPR2020-00614
`
`

`

`ISOC-SNDSS' 
`AnEmpiricalStudyofSecureMPEGVideoTransmissions
`I-frame.TheinclusionoftheI-blocksinP-andB-
`Thecompression,usingspatialandmotiontech-
`framescompromisesthesecurityofthetransmission.
`niques,yieldsabitstreamthatiscompressedeven
`Therearetwopossiblesolutionstoeliminatethisprob-
`furtherbytheencoderusingavariable-length(Hu(cid:11)-
`lem:forcetheencodernottogenerateI-blocksoutside
`man)coding.Itisthecombinationofthequantization
`theI-frame,orencryptallI-blocks.Wewillexplore
`andtherun-lengthencodingthatyieldsthegreatest
`thelatteroptioninalatersection.
`compressioninMPEG.Thisresultingbitstreambe-
`comestheMPEGsequencewiththeadditionofthe
`. PreviousWorkonSecureMPEG
`necessarysequenceandGOPheaders.
`TherehavebeensomeattemptsatsecuringMPEG
`. NaiveEncryption
`transmission.Forexample,MaplesandSpanosper-
`formreal-time,softwareencryptionusingselectiveen-
`Themoststraight-forwardmethodtoachievese-
`codingoftheI-frames.Theencryptionwasperformed
`cureMPEGtransmissionwouldbetoencrypttheen-
`ona-classmachineforATMtransmission[].In
`tireMPEGbit-stream.Oncereceived,thebit-stream
`thiswork,theheadersandtheI-framesintheMPEG
`wouldbedecryptedanddecodedfordisplay.While
`streamareencryptedusingDESencryptionintheci-
`wholesaleencryptionisarguablythemostsecureen-
`pherblockchaining(CBC)mode[].TheCBCmode
`cryptionapproach,selectiveencryptionanddecryp-
`isacommonlyusedoperationmodeofDESandhas
`tion,exploitingtheMPEGstructure,seemstoo(cid:11)er
`theabilitytorecoverpropertyfrombit-errors.
`anencryptionsolutionthatreducesthecomputation
`Inanotherexample,MeyerandGadegasthavede-
`considerablywithoutcompromisingthesecurityofthe
`signedanewMPEG-likebit-streamcalledSECM-
`transmission.
`PEGthatincorporatesselectiveencryption,addi-
`Oneoftheseselectiveencryptionschemesusesthe
`tionalheaderinformationtoaidinbit-errorrecovery,
`factthatthedecoderdependsontheintegrityofthe
`andhashigh-speedsoftwareexecution[].SECMPEG
`I-frametodecodetheP-andB-frames.Encryption
`alsoallowstheuseofbothDESandRSA[]encryp-
`oftheI-framealone,intheory,wouldrenderthein-
`tionalgorithms.
`formationintheP-andB-framesuseless.Encrypting
`SECMPEGimplementsfourlevelsofsecurity.At
`baseframesaloneallowstheencryptiontobeimple-
`the(cid:12)rstlevel,SECMPEGencryptstheheadersfrom
`mentedinsoftwareinreal-time[].Wholesaleencryp-
`thesequencelayerdowntotheslicelayer.Themotion
`tionanddecryptionoftheMPEGbit-streamwould
`vectorsandtheDCT-codedblocksareunaltered.At
`requiremuchfasterprocessingandpossiblydedicated
`thesecondlevelofsecurity,SECMPEGencryptsparts
`hardware.
`oftheI-blocks(presumablytheDCandthelowerAC
`Thisbase-frameencryptionschemedoesnotac-
`termsoftheDCT-encodedI-blocks)inadditiontothe
`countforI-blocks(intracoded(cid:2)blocks)intheP-
`encryptionperformedinthe(cid:12)rstlevel.Atthethird
`andB-frames.Undercertaincircumstancesthese-
`levelofsecurity,SECMPEGencryptsI-framesandall
`curityofthisschemefallsapart.Weperformedone
`I-blocks.Atthefourthlevel,SECMPEGperforms
`experimentwheretextinthesceneundergoestransla-
`wholesaleencryption.
`tion.Theresultingframesequencewasencodedand
`SinceSECMPEGrequiresmajoradditionsand
`theI-frameswereencrypted.However,thetextwas
`changestothestandardMPEGbitstreamheaders,a
`clearlyreadablewhenviewingtheencryptedstream
`SECMPEGbitstreamisnotcompatiblewithastan-
`onastandardMPEGdecoder.
`dardMPEGbitstream.Aspecialencoderandde-
`coderwouldberequiredtoviewunencryptedSECM-
`. E(cid:11)ectsofI-BlocksinP-andB-Frames
`PEGbitstreams.
`TheMPEGstandardallowsencoderstoincludeI-
`blocksintheP-andB-frames.Encoderscanchoose
` WeaknessinSelectiveEncryption
`toencodecertainblocksasI-blockswhentheblock
`cannotbeadequatelymotion-compensated.Framese-
`Asmentionedearlier,oneproposedselectiveen-
`quenceswithalargeamountofmotionoftenrequire
`cryptionalgorithmencryptsonlytheI-framesof
`moreI-blocksinmotion-compensatedframes.
`MPEGvideo[].Thisselectivemethodappearsat-
`tractivebecauseitcansigni(cid:12)cantlyreducetheamount
`IfaframecontainsanunencryptedI-block,blocks
`ofencryptionanddecryption.Thepercentageofre-
`insubsequentframesusingthatblockasareference
`ductiondependsonmanyfactors,suchthefrequency
`wouldbecorrectlydecodeduntilthenextencrypted
`Copyright(c) IEEE.
`
`
`
`DivX, LLC Exhibit 2017
`Page 2017 - 3
`Netflix Inc. et al. v. DivX, LLC, IPR2020-00614
`
`

`

`ISOC-SNDSS' 
`
`AnEmpiricalStudyofSecureMPEGVideoTransmissions
`ofI-frames,theparticularencoderused,andtheac-
`tualvideosequence.
`Inatypicalframesequenceof
`length ,theI-framesandI-blockscantakefrom
`to%ofbitstream,dependingontheamountof
`motioninthesceneandthetypeofencoder.
`Thesecurityrationalbehindthisproposalisthat
`sinceI-framescarrythebasicinformation,iftheyare
`unavailabletoaneavesdropper,thentheP-andB-
`framesbecomeuselesseventhoughtheyaretransmit-
`tedunencrypted.Byrestrictingtheencryptionand
`decryptiontotheI-frames,thismethodreducesthe
`cryptographiccomputationbyhalf.Thisreductionis
`signi(cid:12)cantbecausethedecryptionoftheI-framesand
`I-blockscantakeasmuchasone-thirdofthedecoding
`time[].However,alittlecryptanalysis(e.g.,[ ])wouldre-
`vealthat,althoughasingleP-orB-frameonitsownis
`meaninglesswithoutthecorrespondingI-frame,ase-
`riesofP-orB-framescarriesmuchinformationiftheir
`baseframesarecorrelated.Forexample,inaMPEG
`sequenceofanongoingmeeting,thebaseframes(the
`I-frames)oftencontainimagesofthesamepeoplein
`thesamesurrounding,whichareclearlyrelated.Thus
`weexpectthatforsuchMPEGsequencesthesimple
`selectiveencryptionmethod[,]wouldnotbeable
`toprovideahighlevelofsecurity.
`Toverifyourhypothesis,weconductedaseriesof
`experiments.Theexperimentenvironmentandsetup
`isasfollows.Thestructureofourexperimentalsys-
`temisasshowninFigure .Wewrotetheencryption
`routinethatencryptsvideooutputfromaMPEGen-
`coderandmodi(cid:12)edtheMPEGencodercodeaccord-
`ingly.Morespeci(cid:12)callyweusedtheparallelMPEG-
`encoderdevelopedatSRI[ ]andtheMPEG-encoder
`byChadFogg,usedDESincipher-feedbackmodeto
`generateacipherstream,andexclusive-orthecipher
`streamwiththeMPEGstream. Theroutinecanop-
`tionallyencryptI-framesonly,orincludingallI-blocks
`also,orallP-framesinadditiontoallI-blocks.
`Forconvenience,thisstudywouldneedavideode-
`codertoplayencryptedvideosequences,andwede-
`cidedtouseo(cid:11)-the-shelfproductstokeepcustomiza-
`tionataminimum.Thisdecisionbringscertaincon-
`straints.Forexample,wecannoteasilyexperiment
`withthee(cid:11)ectofencryptingheaderinformation.
`
`

`

`ISOC-SNDSS' 
`AnEmpiricalStudyofSecureMPEGVideoTransmissions
`andterracehouses.The\(cid:13)owers"videocontainsob-
`jectinthescenethatmoveinbothdirectionssimul-
`taneously.
`
`Figure:MissAmericaI:unencrypted
`Wethenmodi(cid:12)edtheMPEGencodertoencrypt
`theI-framesandplayedtheciphertextdirectlyona
`MPEGplayer.Becauseoftheinterframecorrelation,
`wecanclearlyseethatthevideosequencecontainsa
`personspeakingandcanalsomakeoutsomeofthe
`featuresoftheperson.Arevealingframefromthe
`encryptedsequenceisshowninFigure.
`
`Figure:FlowersI:unencrypted
`WeencryptedtheI-framesinthissequenceand
`playedtheencryptedstreamdirectlyontheVMPEG
`player.Figureclearlyshowsthatpartsofthatframe
`aree(cid:11)ectivelyinplainview.Greatportionsofthe
`videoarevisiblepartlyduetointerframecorrelation
`andmainlyfromunencryptedI-blocksintheP-and
`B-frames.Therefore,encryptingI-framesonlymay
`leavevisibleI-blocksinotherframes.
`
`Figure:MissAmericaII:\encrypted"
`Thelackofsecuritycaninfactbemuchworseun-
`dersomecircumstances,forexample,ifthereisagreat
`dealofmotioninthescene.Thenextvideoexample
`illustrateshowtheselectiveencryptionschemecanfall
`apart.Figureshowsaframefromthe\(cid:13)owers"video
`sequencetakenonamovingplatformatthebottom
`ofaslopepanningacrossascenethatcontainshouses
`anda(cid:13)owergarden.Thee(cid:11)ectappearstobetrees
`movingtowardsleftagainstabackgroundof(cid:13)owers
`Copyright(c) IEEE.
`
`
`
`Figure:FlowersII:\encrypted"
`Thesetwoexampleshavedemonstratedclearly
`that,contrarytosomepreviousproposals,MPEG
`videocannotbesecuredbysimplyencryptingtheen-
`codedI-frames.Ofcourse,securitymeasureislargely
`amatterofthesystemrequirements.So,simpleselec-
`tiveencryptionmaybesu(cid:14)cientforcertainapplica-
`tions,suchascabletelevisionbroadcastofpaychan-
`nels,wherequalityofservicesupersedeshighsecurity
`asthedriver.However,wefeelthatawiderangeof
`
`DivX, LLC Exhibit 2017
`Page 2017 - 5
`Netflix Inc. et al. v. DivX, LLC, IPR2020-00614
`
`

`

`ISOC-SNDSS' 
`
`AnEmpiricalStudyofSecureMPEGVideoTransmissions
`applicationswouldrequireamuchhigherlevelofse-
`curityforMPEGvideo.Forexample,supposeaper-
`soninCupertino,California,isgivingacon(cid:12)dential
`presentationonAppleComputer's(cid:12)nances,whichis
`transmittedtoagroupofinvestorsinNewYorkCity
`throughtheInternetusingMPEGencoding.UsingI-
`frameselectiveencryption,itislikelythatsomepor-
`tionsofthetextusedinthepresentation(e.g.,dollar
`numbersprojectedoverhead)wouldbevisibletoan
`eavesdropper,andthislevelofsecurityisclearlyun-
`acceptable.
` Tradeo(cid:11)sinEncryptingMPEG
`Inthissection,wesuggestafewmethodstoim-
`proveMPEGsecurityanddiscusstrade-o(cid:11)sbetween
`performanceandthelevelofsecurity.
`. ChangingI-FrameFrequency
`AsimplesecurityimprovementinselectiveMPEG
`encryptionistoincreasethefrequencyoftheI-frames.
`TheincreaseinencryptedI-frameswouldreducethe
`interframecorrelation.Atthemaximumfrequency
`theentirevideosequenceiscomposedonlyofI-frames
`resultingintheencryptionoftheentiresequence.This
`schemeisobviouslyquitesecurebutcompressionand
`speedwouldbemuchdegraded.Selectiveencryption
`isintendedtoaddsecuritywithoutreducingthecom-
`pressionenjoyedbyusingP-andB-frames.
`WeencryptedafewsequenceswithincreasingI-
`framefrequencies.Byviewingtheseencryptedvideos,
`weobservedthatastheI-framefrequencyincreases
`theamountofvisibleinformationdecreases.Wealso
`noticedthatoftentheencryptedsequenceappears
`quitesecurewellbeforetheI-framefrequencyreaches
`maximum.Forexample,asthefrequencyofI-frames
`increasesto inevery framesinencodingthe\Flow-
`ers",wefoundthatthemostrevealingframelookslike
`theoneshowninFigure.
`Heretheframelooksrandomexceptcertainsmall
`areas(e.g.,treebranchesintheupper-leftcornerand
`(cid:13)owersinthelower-rightcorner)appearuna(cid:11)ected
`becausewedidnotencrypttheI-blocks.
`Asexpected,whenweincreasedtheI-framefre-
`quencytothemaximum,thewholesequenceappeared
`random.Thecaseisthesamewiththe\MissAmer-
`ica"example,asshowninFigure .
`Basedonourexperiments,wefeelthatitisprob-
`ablyinfeasibletocomeupwithasetofparameters
`thatwillworkreasonablywellacrossawiderangeof
`di(cid:11)erenttypesofvideothatarecommoninvarious
`Copyright(c) IEEE.
`
`
`
`Figure:FlowersIII:higherI-framefrequency
`applicationsusingMPEG.Thusitwouldbedesir-
`abletobuildintoMPEGencryptionfacilityacontrol
`mechanismtodynamicallychangeI-framefrequency
`basedonvideotypes.
`Itwouldalsobeveryuseful
`toconductfurtherexperiments,usingextensivevideo
`footage,toobtainforvarioustypesofvideostreams
`referencepoints(thatcorrespondtodi(cid:11)erentsecurity
`levels)andperformancebenchmarks.Forexample,a
`lowlevelofencryptionmaybesu(cid:14)cientforscrambling
`videosofice-hockeymatches,becauseoftherapidly
`changingbackgroundandfastmovingplayers.On
`theotherhand,asnookermatch(pool,inAmerican
`terms)mayrequireadi(cid:11)erentsetofparametersbe-
`causethesnookertableisoftenshotfromafew(cid:12)xed
`anglesandtypicallyonlyoneortwoballswouldmove
`atonetime.Thesedatacouldguidefurthersecure
`MPEGdevelopmentand(cid:12)netuningofapplication-
`speci(cid:12)cMPEGsecurityparameters.
`
`Figure :MissAmericaIV:fullyencrypted
`
`DivX, LLC Exhibit 2017
`Page 2017 - 6
`Netflix Inc. et al. v. DivX, LLC, IPR2020-00614
`
`

`

`ISOC-SNDSS' 
`AnEmpiricalStudyofSecureMPEGVideoTransmissions
`. EncryptingI-BlocksandP-Frames
`ofthenatureofMPEG,alsoa(cid:11)ectsthee(cid:11)ectiveness
`ofvideocompression(thusincreasingthetotalsizeof
`theMPEGbit-stream).Forexample,inthe\Flowers"
`Tofurtherimprovesecurity,weexperimentedwith
`sequence,increasingfrom inevery framesto in
`encryptingallofI-blocksintheP-andB-framesin
`every nearlydoublesthevideosize.Tablebelow
`additiontoencryptingallI-frames.Wefoundthat
`detailsthechangesinvideosizeswithdi(cid:11)erentI-frame
`thisincreasessecuritysigni(cid:12)cantly,makingitpossible
`frequencies.
`toreducetheI-framefrequencyandimprovingcom-
`pressionwhilemaintainingthesamelevelofsecurity.
`Figure showsthatbyencryptingI-blocks,thefol-
`lowingqualityofencryptionisobtainedeventhough
`Table:Videosizes(inbytes)atdi(cid:11)erentI-frame
`theI-framefrequencyisonly inevery.
`frequencies
`
`Figure :FlowersIV:I-blocksencrypted
`Anotherimprovementweexperimentedwastoen-
`crypttheP-frames.Ourintuitionwasthatencrypting
`theP-framesismoreeconomicalthanincreasingthe
`encryptedI-framefrequencybecausetheformerdoes
`notcausethevideosizetoincrease.Bothmethods
`wouldincreasethecomputationload,althoughthefor-
`merislesscostlybecauseitresultsinsmallerencoded
`streams.Weencryptedaseriesofvideosequences,withdif-
`ferentI-framefrequenciesanddi(cid:11)erentencryptionop-
`tions.WefoundthatencryptingtheP-framesdoesnot
`appeartoaddsigni(cid:12)cantsecurityvalue,ifwealready
`encryptallI-blocksinallframes.Wesuspectthatthis
`observationmightbeduetootherreasons,including
`theprecisionandsensitivityoftheobservationinstru-
`ments(inourcase,oureyes)andtheperformanceof
`thevideoplayerweused.Morepowerfulcryptanaly-
`sistoolsmaydiscoverimprovementinsecuritylevels
`whenP-framesareencrypted,andfurtherstudyin
`thisdirectionisnecessary.
`. Trade-O(cid:11)s
`IncreasingtheI-framefrequencynotonlymeans
`thattherearemoreframestoencryptbut,because
`
`

`

`ISOC-SNDSS' 
`AnEmpiricalStudyofSecureMPEGVideoTransmissions
`theACM, ():{,April .
`timesI-blocksalso)aretobeencrypted,hasbeenpro-
`posedtobeaneconomicalandsecuresubstitutefor
`[ ]A.G.Konheim.Cryptography:APrimer.John
`fullencryption.
`Wiley&Sons,Inc.,NewYork,  .
`Inthispaperwehavereportedanempiricalstudyof
`MPEGvideoencryption.Wefoundthatthesemeth-
`[]BenoitM.MacqandJean-JaquesQuisquater.
`odsarenotadequateforsensitiveapplications.Specif-
`CryptologyforDigitalTVBroadcasting.Proceed-
`ically,ourexperimentscon(cid:12)rmedourintuitionthat
`ingsoftheIEEE, (): { ,June .
`encryptingI-framesalonemaynotbesu(cid:14)cientlyse-
`[]T.B.MaplesandG.A.Spanos.PerformanceStudy
`cureforsometypesofvideo.Wehavegivenexamples
`ofaSelectiveEncryptionSchemefortheSecurity
`wherebyplayingback\encrypted"videosequences
`ofNetworked,Real-TimeVideo.InProceedingsof
`onregularMPEGplayers,onecanclearlyseepat-
`thInternationalConferenceonComputerCom-
`ternsofmovementandsometimesevenlargechunks
`municationsandNetworks,LasVegas,Nevada,
`ofplaintextvideo.Weconductedexperimentswith
`September .
`bothMPEG- andMPEG-andobtainedsimilarre-
`sults.Thereareseveralobviouswaysofimprovement.
`[]J.MeyerandF.Gadegast.SecurityMechanisms
`forMultimedia-DatawiththeExampleMPEG-I-
`First,byincreasingtheI-framefrequency,betterse-
`Video.ProjectdescriptionofSECMPEG,Techni-
`curitycanbeobtainedbutattheexpenseofincreased
`calUniversityofBerlin,Germany,May .
`videosize,ahighernetworkbandwidthconsumption,
`andhighercomputationalcomplexityforencryption
`[]R.L.Rivest,A.Shamir,andL.Adleman. A
`anddecryption.Moreover,encryptingtheI-blocksin
`MethodforObtainingDigitalSignaturesand
`allframesimprovessecurity,doesnota(cid:11)ecttheband-
`Public-KeyCryptosystems.Communicationsof
`widthconsumption,andshouldbeconsideredanalter-
`theACM, (): { ,February .
`nativetoincreasingtheI-framefrequency.Additional
`[]DataEncryptionStandard.
`(U.S.)NationalBu-
`encryptionoftheP-framesdoesnotseemtoimprove
`reauofStandards,January .
`(U.S.)Fed-
`security,butwarrantsfurtherinvestigation.
`eralInformationProcessingStandardsPublica-
`Thereareanumberofinterestingissuesforfurther
`tion,FIPSPUB.
`study.Oneistoinvestigateotherwaysinwhicha
`videostreamcanbebrokenintoframesorlayers,and
`theapplicabilityandthesecurityofselectiveencryp-
`tioninsuchsituations.Anotherusefule(cid:11)ortistocol-
`lectdataonawiderangeofvideosamplesandobtain
`approximateparametersfore(cid:11)ectiveande(cid:14)cienten-
`cryption.Analyticalresultswillprovidedeeperinsight
`intothe(in)securityofselectiveencryption.Finally,
`itmaybeworthwhiletoexaminethedi(cid:11)erentorders
`inwhichcompression,errorcorrectioncoding,anden-
`cryptionareperformed,andseeifsomearrangements
`substantiallyimprovesecurityandperformance.
`References
`[ ]IskenderAgiandR.Jagannathan.APortable
`Fault-TolerantParallelSoftwareMPEG- En-
`coder.
`InB.FurhtandM.H.Hamza,editors,
`ProceedingsoftheSecondIASTED/ISMMIn-
`ternationalConferenceonDistributedMultimedia
`SystemsandApplications,Stanford,CA,August
` .
`[]D.LeGall.MPEG:Avideocompressionstandard
`formultimediaapplications.Communicationsof
`Copyright(c) IEEE.
`
`
`
`DivX, LLC Exhibit 2017
`Page 2017 - 8
`Netflix Inc. et al. v. DivX, LLC, IPR2020-00614
`
`