DOC

Authentication

By Marie Martin,2014-06-20 05:39
9 views 0
Authentication

    Authentication

    J.Cent.SouthUniv.Techno1.f2007)04056305

    DOI:10.1007/s117710070108Y

    Authenticationbasedonfeatureofhandwrittensignature

    ZHUShuren(~树人

    Springer

    (1.SchoolofInformationScience,GuangdongUniversityofBusinessStudies,Guangzhou510320,China;

    2.SchoolofComputer,BeijingUniversityofAeronauticsandAstronautics,Beijing100083,China)

    Abstract:Thetypica1featuresofthecoordinateandthecurvatureaswel1astherecordedtimeinformationwereanalyzedinthe

    handwrittensignatures.Inthehand

    writtensignatureprocessl0biometricfeaturesweresummarized:theamountofzerospeedin directionanddirectionv,theamountofzeroaccelerationindirectionanddirectionY,thetota1timeofthehandwrittensignatures,

    thetota1distanceofthepentravelinginthehand

    writtenprocess.thefrequencyforliftingthepen,thetimeforliftingthepen,the amountofthepressurehigheror1owerthanthethresholdvalues.Theformulaeofbiometricfc:aturesextractionweresummarized.

    TheGaussfunctionwasusedtodrawthetypicalinformationfromtheabove

    mentionedbiometricfeatures,withwhichtoestablish

    thehiddenMarkovmodeandtotrainit.Theflameofdoubleauthenticationwasproposedbycombingthesignaturewiththedigital

    signature.Webservicetechnologywasappliedinthesystemtoensurethesecurityofdatatransmission.Thetrainingpractice

    indicatesthatthehand

writtensignatureverificationcansatisfytheneedsfromtheofficeautomationsystems.

    Keywords:behavioralbiostatisticsfeature;hand

    writtensignature;hiddenMarkovmode;signatureverification 1IntrOductiOn

    Personalauthenticationisbecomingnecessaryin moreandmorefields.Thetraditionalpersonal authenticationmethodscannotkeepupwiththe developmentofthesocietybecauseoftheirinherent defects.Undersuchcircumstance,biometricsemergedas thetimerequires.Biometricsisthetechnologiesthat analyzehumancharacteristicsforautomatedpersonal authentication.Onlinehandwrittensignatureverification isanimportantbranchofbiometrics.Handwritten signaturehasbeenahumanbehaviorcharacteristicand beenwidelyappliedsinceancienttimes.Itrepresents

    theuser'sintrinsicanduniquetraits.andisalsorelatedto thespecificbiomechanicalsystem.Thebehavioral

    biostatisticst6Jisexpressedbythelongtermbehavioral

    characteristicsfthehandwrittensignatures,thecontrail ofpivotalstrokes)Thehandwrittensignature

    verificationisbasedontheconceptsthateveryone's signatureconveyshisuniqueunderstandingandthe modeofwritingconformstothebehavioral

    biostatistics[.Thehandwrittensignatureverification systemhasbecomeahottopicofidentityverification technology.

    Someresearchersconsideredcommonissueswith extractionofidentificationdatafromvarioustypesof biometrics,andprotectionofsuchdataagainst

conceivableattacks[(引.Thevaimedatfacilitating

    reliablebiometricsauthenticationbyimproving authenticationprecisenessandprovidingcounter

    measuresagainstattackstoanauthenticationsystem.In thisarticle,thetypicalfeaturesofthecoordinateandthe curvatureaswellasthetimeinformationrecordedinthe handwrittensignatureswereanalysed,the10biometric featuresinvolvinginthehandwrittensignatureprocess

    wereobtained.TheGaUSSfunctionwasusedtodrawthe typicalinformationfromtheabovementionedbiometric

    features,withwhichtoestablishthehiddenmarkov mode(HMM)andtotrainit.

    2Modelingofhand-writtensignature

    verificationsystem

    2.1Potentialfe-atures

    Thefeatureextractionisthekeystepinthe

    recognitionoftheon..1inehand..writtenChinese characters[.thequalityofthefeatureextractioncan directlyaffecttherecognitionperformance.The coordinatefeatureextractionandthecurvaturefeature extractionwereusedinthepaper.TheCOOrdinatefeature extractionregardsthecoordinateatthespecialspotsas thestrokes'features.ittakespositioninformationatthe specialspotsintoconsiderationandexpressesthe directionalinfc)rmationatthespecialspotsinthestrokes. Thecurvaturefeatureextractionregardsthecurvatureat thespecialspotsasthestrokes'features;itfoCHSeSonthe Foundationitem:Project(03llY31021supportedbytheNaturalScienceFoundationofHuna

    nProvince,China

Receiveddate:2006——12——24;Accepteddate:2007——03——27

    Correspondingauthor:ZHUShuren,PhD,Professor;Tel:+8615913137689;E

    mail:zhusr@gdcc.educn

    564J.Cent.SouthUniv.Techno1.2007,14(4) movementinformationofthestrokes'figure.According tothefeamresofthecoordinateandthecurvature,and therecordedtimeinformation,aseriesofbiological featuresofthesignaturesrecognizedwereobtainedby system.10primarybiometricfeaturesweresummarized asfollows:

    1)?listheamountofzerospeedindirection; 2)?1istheamountofzerospeedindirectionY(the absolutezeroisspanless,sothezeroisreplacedbya speedvaluethatisalittlelsmallerthanpeakvalue); 3)?istheamountofzeroaccelerationindirection x;

    4)istheamountofzeroaccelerationindirection y(theamountofzeroexpressesthemovementofspeedup tospeed--downorspeed--downtospeedup); 5)twisthetotaltimeofthehand-writtensignature; 6)Distotaldistanceofthepentravelledonthe hand-written,namelytheEucliddistanceofallthe points:

    D=?{(—川)+(y),0,1,,?_1

    (1)

    wherexisthecoordinateindirectionxandYisthe coordinateindirection):

    7)?wisthefrequencyofliftingthepen;

    8)tnisthetotaltimeofliftingthepen(namelythe

    totaltimefromthebeginningtotheendingofthe writing);

    9)Nmaxistheamountofpressurehigherthanthe thresholdvalueTmax(accordingtothehardwarefeatures ofthehand-writtenpen);

    10)Nministheamountofpressurelowerthanthe thresholdvalueTmin(accordingtothehardwarefeatures ofthehand.writtenpen).

    2.2Featureextraction

    Thesystemdrawssuchfeaturesasthetotaltime, thetotallengthofthestrokesandtheamountofzero speedfromthesignatures,andthenconstructsamodel f0rthesefeatureswiththeGaussfunctionthatcan computetheprobabilitydensity,namelythroughdensity functionitcanestimatealltheparametersofthefeatures fromthesignatureswatches.Intheestimationprocess itobtainsthecorrespondingaverageandvarianceofthe features,andregardsthemastheuniquesignsofthe signature.

    Intheprocessofdesignationandrealizationofthe system,itisdiscoveredthatthesignatureisclosely connectedwiththeuser'spsychology.Andthesignature hasagreatdifferenceindifferentsurroundingsandtime; andsomeotherfactorscanalsoexertgreatinfluenceson it.Forexample,thetotaltime,thelengthofthestrokes, thetimeforliftingthepencanbedifferentforthesame user'sdifferentsignatures.Thesefeaturesoscillateabout theaverageandvariancewhichcansymbolizeone person'sbiometricfeatures{.Thereforeitcanbe consideredasaproperchoicetodescribethesefeatures

    throughthedistributionofacertainkindofsignatures andtoverifythesefeaturesthroughexperimentswiththe Gaussfunction.

    Thefollowingpartoffersadetailillustrationto theseformulaehowtodrawthefeatures.

    SpeedVxandv,expressthefunctionsoftime, whichcanbecalculatedwiththefollowingformulae: IV=(,+1)/(f+1IV:(+1)/r,+1(t)

    (

    t)

    wheremistheserialnumber,m=O,1,,?1;xis

    thecoordinateindirection;YistheCO0rdinatein direction;t?isthetimeofmovement;Vxisthespeedat pointtindirection;V,isthespeedatpointtmin directionY.

    Accelerationaranda|canbecalculatedwith differentspeeds,theaccelerationatpointtindirections xandYcanbeexpressedinthefollowingformulae: I=}+lV/r,+1

    a=v3.j/(1m+I

    

    t

    

    tmj

    whereaxistheaccelerationatpointtindirectionx,ay istheaccelerationatpointtmindirectionv. Theaveragedescribesthegeneraloftherecorded datacollectivity:thevariancedescribesthedistributing patternofeachgroupinthefeaturespace.TheNgroups ofsignatureswatchescandeduceNvaluesofonefeature,

    throughtheGaussfunctionthesevaluescansupplythe averageestimateandvarianceestimatethatcanshowthe distributionofparameterizeddefinitions.The10 biologicalfeaturescanbededucedbytheformulaeinthe followinganalysis.

    Theagonicestimate)oftheaveragecanbe

    illustratedinthisway:

    k=x1

    1

    wherexshowsthe10biologicalfeatures.

    Theagonicestimate()ofthevariance

    illustratedinthisway:

    canbe

    ?一1

    =

    ?(-/4)/(N1)(5)

    1

    wherexishowstheabove-mentioned10biological features,expressestheagonicestimateofthe divideddifference;kistheamountoffeatures,which

    ZHUShuren,etal:Authenticationbasedonfeatureofhandwrittensignature565 meansthatVkisthevalueofkfeatures.Therefore,,) candescribethefeatureofeachuser.Thesystemthat collectsthe10featurescandeducethefollowingfigures like(1,VL),(/-/2,),,10,V10).Thereferredsignature templateofeachsignatorythatisusedforauthentication canbebuiltwiththosevectors.

    Iftheuserhasfoursignatureswatches,onlytwoof themaretakenintoconsiderationinthefollowing

patterns:thetota1timeofthehandwrittensignatureand

    thefrequencytoliftthepen.Signature1:4.234s.5 times;Signature2:4.623s.6times;Signature3:4.555s, 4times;Signature4:4.772s.5times.Forthetotaltime. theaverageis4.546s.thevarianceis0.051477;forthe frequencytoliftthepen,theaverageis5,thevarianceis 0.666.Sotheextractedfeaturesofthefoursignature swatchesare(4.546,0.051477),(5,0.666). Thesignaturecanberegardedasatwodimensiona1

    curvethattakes1,Y2)(0,0)asthebeginningand?'

    YN)=(,V/y)astheend,theshortestofthelengthis: ?一l

    D(,)=?((Xk1,Yg1),(,y))(6)

    'k=2

    whereDistheshortestlengthofthecurve,Xkisthe coordinateindirectionx,Ykisthecoordinateindirection .

    Theinformationofspecia1pointsxandvobtained fromtheabovecontro1canbeusedtomodulatethe abovementionedformulae;thenthefollowingformula canbededuced:

    ?一l

    ?一Xi+1)+(Yi+t)

    l

    whereDtisthetotallengthofthecurve

    2.3Signatureverification

    2.3.1HiddenMarkovmodel

    HiddenMarkovmodel(HMM)includesaMarkov chain{t:1>p0}andanobservedstochastic processy_-:1>p0},whereintheMarkovchainis

    unobservable,butcanonlyberecognizedthroughan observedstochasticprocess[.2..TheparametersetOf oneHMMisasfollows.

    1)Themodelstatecollectionis{l,,,};

    2)TheobservationstatecollectionV--{,,,

    },forthesuccessiona1observedquantitiesthe collectionisinfinitudes;

    3)Theinitialprobabilityis{},therein7r

    P(ql=),1???;

    4)ThetransferprobabilityisA={a0},wherein

    P(qf_1S,.Iq,Si),1?f,J?_?;

    5Instatethevisibleprobabilitydistributionofthe signisB={6,(},thereinbj(k)=P(attimepointtthe signislg,:),1???,1??

    2.3.2HiddenMarkovclassifier

    InthesignatureverificationprocesstheHMMis usedasafeatureclassifier.ThescatteredHMMis expressedby5parameters:modelstate,observationsign, initia1probabilitY,transferprobabilityandemission probability.Themode1stateandobservationsignare differentinaccordancewiththeirapplication.The scatteredHMMclassifierisbuiltforeachsignature.and iscomposedofstrokestypeclassifierHMM1andstrokes positionclassifierHMM2.

    InstrokestypeclassierHMM1.thenumberofthe elementsintheobservationsetexpressestheamountof thesignaturestrokes;eachobservedsignshowsonekind biologica1features.forthe10kindsbiologicalfeatures theamountofobservedsignsis10:theamountofmode1 statesis20.TheimputofHMMisthebiologicalfeatures

    obtainedfromtheinitia1featurevalues.thesignatureis modeledinaccordancewiththetypesofstrokes. InstrokespositionclassifierHMM2.themargin betweenthenumberoftheelementsintheobservation setandtheamountofsignaturestrokesis1.each observedsignexpressesonekindofstrokes'position.for 9kindsofstrokes'positiontheamountofobservedsigns is9:theamountofelementsinthemode1statesetiS20; thepositionconnectionisdeterminedbythedirection vectorthatpointsarefromtheendingspot(tothe

    beginningspot(.Twocategoriesareclassifiedin accordancewiththecoincidenceofEandS.Ifthereisno coincidencebet,veenEandS.8strokes'positionscanbe deduced:l(East),2(Eastnorth),3(North),4(Westnorth), 5(West),6(Westsouth),7(South),8(Eastsouth);ifthere isacoincidencebetweenEandthereonlyonestrokes' positioncanbededuced.

    HMMcombinesHMM1withHMM2.thenumber

    ofelementsineachmodel'sobservationsetreferstothe amountofsignaturestrokes;eachobservedsign expressesonekindofbiologica1featureoronekindof strokes'position:theamountofobservedsignsis19. thenumberofelementsinmodelstatesetiS20.

    2.4Verificationmodel

    Thesignatureverificationsystemcouldcertainlybe insertedintosomeapplicationa1systemconvenientlyand efficiently,soitplaysitsrolecompletely'15J.Inorder toprovideamoreagileandconvenientmethodwiththe combinationofeveryofficeautomation(OA)system,the portionofsignatureverificationcanbesuppliedthrough

Report this document

For any questions or suggestions please email
cust-service@docsford.com