Availableonlineatwww.sciencedirect.com
Journal of Applied Research and Technology
www.jart.ccadet.unam.mx JournalofAppliedResearchandTechnology15(2017)402–411
Original
De-noising and spoofing extraction from position solution using wavelet transform on stationary single-frequency GPS receiver in immediate
detection condition
Mohammad Reza Mosavi
∗, Amir Reza Baziar, Maryam Moazedi
DepartmentofElectricalEngineering,IranUniversityofScienceandTechnology,Narmak,Tehran16846-13114,Iran Received25July2016;accepted7April2017
Availableonline26August2017
Abstract
Thegrowingdependenceofcriticalcivilinfrastructureonglobalpositioningsystem(GPS)makesGPSinterferencenotonlyasafetythreat, butalsoamatterofnationalsecurity.TheresearchdoneinthispaperisinitiatedbytheneedtodiminishthistroubleonGPSbasedpositioning.
Thesuggestedcompensationtechniqueassumesthatthepresenceofaspoofingsignalisimmediatelydetermined.Thepositionresidualsofthe lastauthenticandnewfakesignalsarepassedtothewavelettransform(WT).WeutilizedWTforde-noising.Afterwards,positiondeviationsdue toanattackcanbeextractedandthentheestimatedpositionofthereceivedsignalwillbecorrected.Asaprimarystep,theproposedalgorithm hasbeenimplementedinastationarysoftwareGPSreceivertoprovetheconceptoftheidea.Theperformanceofthetechniqueisvalidatedusing severallaboratoryandmeasurementdatasets.Interferencemitigationhavingtoleranceof3%andaverageof99.5%isyieldedonlaboratorydata setandcompletecompensationisachievedonmeasurementdataset.Thetestresultsshowthattheproposedtechniquesupremelygainstrength thereliabilityofcivilstationaryGPSreceiveragainstinterference.
©2017UniversidadNacionalAutónomadeMéxico,CentrodeCienciasAplicadasyDesarrolloTecnológico.Thisisanopenaccessarticleunder theCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Keywords: De-noising;Spoofingextraction;GPSreceiver;Wavelettransform
1. Introduction
Fromcarstocommercialairplanes,globalpositioningsys- tem (GPS) technology isubiquitous and it canbe hacked or
“spoofed”.Inspoofingattack,thecounterfeitGPSsignalisgen- eratedtomanipulateatargetreceiver’spositionortime.Spoofing attackcanbeclassifiedintothreemaingroups:simplistic,inter- mediateandsophisticatedattacks. Simplisticspooferattaches a poweramplifier and an antenna to aGPS signalsimulator andthen radiates the radiofrequency(RF) signal towardthe targetreceiver.The secondgroupsynchronizes itscounterfeit signalswiththeauthenticGPSsignals.Therefore,thefakesig- nalscanmore-easilybemasqueradedasgenuine.Sophisticated
∗Correspondingauthor.
E-mailaddress:m[email protected](M.R.Mosavi).
PeerReviewundertheresponsibilityofUniversidadNacionalAutónomade México.
attackscontainseveralreceiver-spoofersutilizingsharedrefer- enceoscillatorandacommunicationlinkadjustedtothetarget antenna. SimplisticspoofercanproduceGPS signal,butcan- not make themtobealigned withthecurrent broadcast GPS signals.However,iftheadversarycantransmitsignalswiththe powerhigherthanthatofthelegitimatesignal,misleadingcom- mercialreceiverswouldbepossible.Thethirdcategoryisthe most effective method of the spoofing generation.Whatever, physicallimitationsforplacingtheattackerantennatowardthe victim receiverhave made their implementation so hardand impossible insomecasesbecause of targetreceiver’s motion (Baziar, Moazedi, & Mosavi, 2015). Although, the receiver- spoofercanbeformedsmallenoughtoplaceindistinctlynear theantennaofthevictimreceiver.Itseemsthattheapplicable versionofspoofingwillbetheintermediatesinceitisaccessible andimplantableinsoftware-defined-receiver(SDR).Therefore, we will oppose an intermediate spoofing in which the main GPS signalisre-senttothetargetreceiveraftersomeprecise delay.
http://dx.doi.org/10.1016/j.jart.2017.04.001
1665-6423/©2017UniversidadNacionalAutónomadeMéxico,CentrodeCienciasAplicadasyDesarrolloTecnológico.Thisisanopenaccessarticleunderthe CCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Duringthepastdecade,severalanti-spoofingalgorithmshave beendevelopedandtested(Lee,Kwon,An,&Shim,2015).Of course,spoofrecognitionhasbeenmoreoutstanding.Adesir- ablecriterionforcounteringalgorithmsisthatitcanbeadded tocivilGPSreceivers easily.Throughmanysuggestedspoof- recognition methods, a real-time detectionalgorithm in GPS receiverisaccessible.However,the necessityof developinga newspoofmitigationalgorithmthatcanbeimplementedsimply incivilGPSreceiverisclearlyobservable.
In this paper, we have introduced a novel technique for GPSdataprocessingbasedonwavelettransform(WT),which assumesthatareal-timespoofdetectionmethodstandsonSDR.
Inordertoimplementsimplyinthereceiver,thesuggestedalgo- rithmhasbeenimplementedinthepositionsolutionoftheGPS receiver. In our anti-spoofingapproach, we also performed a de-noisingprocedurebasedstationaryWT(SWT)overposition residuals.
Therestofthispaperorganizedasfollows.Section2reviews previousanti-forgeryapproacheswiththeiradvantagesanddis- advantages. Section 3 proffers a novel spoof compensation algorithm;afterashortdescriptionofmainidea,wewillintro- duceWTanditsapplicationprinciplesinGPSsignalprocessing.
Thissectionthenpresentsadetailedexplanationoftheimple- mented algorithm. Section 4 discusses both laboratory and measurementspoofingdatasetsandtestresultstakenfrominter- ferencereduction.Aprecisecomparisonbetweenpreviousand suggestedtechniques iscoming up atthe endof the section.
Finally,Section5statestheconclusion.
2. Previouslyproposedinterferencereductionmethods Mostoftheattemptsinanti-spoofingfielddedicatetodetect andfewof themrelate tospoofmitigate.Sinceoursubjectis spoofreduction,wedonotstudyrecognizingmethodsbutitis presentedareviewofspoofreductiontechniqueswithafault- findinglookonthem.
The proposedmethod inLin, Haibin, and Naitong(2007) constantlyinvestigatestheinternalandexternalinformationand estimatetheauthenticsignal. Thisreceiverhastwo operation modes.Inthenormalmode,thereceiverreliesonthecollected informationandin the alertmode, it compares the predicted valuesagainsttotheobtainedposition-velocity-timesolutions.
ThelocationpredictioninthissystemisimplementedbyKalman filteringor inertial sensors.Thisalgorithm isnot suitable for long-timespoofsbecausetheestimationerrorgrowsduringthe attack.
InMosavi,Nasrpooya,andMoazedi(2016),theadaptivefil- terisusedforestimatingtheparametersofauthenticandforgery signals.Interferenceeliminationisperformedbysubtractingthe estimatedconflicteffectsfromthemeasuredcorrelationfunc- tion.
Due to practical limitations, simplistic and intermediate attackstransformseveralcounterfeitsignalsfromsingle-source, whilethelegacysignalsaretransmittedfromdifferentsatellites anddirections.Thus,spatialprocessingcanbeusedtoestimate thethree-dimensionaleffectsofthereceivedsignals(Magiera&
Katulski,2015)byantennaarray.Therearethreedifferentways
to implement the antenna array. First, multi-antenna receiver utilizesanarrayprocessingtechniquetoshapeitsbeam.After denotationofcounterfeitsignaldirection,itsteersanulltoward the attacker source and so negates harmful influence. The two-antenna array detects the different incident signals from different antennasby cross-correlation.Moreover,bymoving ahandheldreceiverwithasingle-antenna,aformofasynthetic array can be generated (Broumandan, Jafarnia-Jahromi, &
Lachapelle, 2015).This techniqueis reliable,but takesmore computationalandhardwarecomplexitysoitisnotimplantable incommoncivilreceivers.Themainideainvector-based(VB) trackingtechnique iscombiningbothnavigationsolutionand thesignaltrackingburdensinordertoincreasetherobustness ofGPSreceiversagainsttotheinterference(Jafarnia-Jahromi, Lin, Broumandan, Nielsen,& Lachapelle,2012).It increases thecomplexityofhardwareandprocessing.
Another technique, called receiver autonomous integrity monitoring (RAIM),usesthe redundantinformationtodetect andmitigateintegritythreatatthenavigationandpositionsolu- tion level (Ledvina, Bencze, Galusha, & Miller, 2010). This methoddetectsdamagedpseudo-rangesorcarrierDopplershift frequencyandexcludesthe measurementerrorsfromnaviga- tion solution via statistical hypothesis testing. It is effective onlyincaseswhereonlyoneortwospoofedmeasurementsare presentamongseveralauthenticpseudo-ranges.They arealso quiteeffective fortheless sophisticatedattacks.It seemsthat theGPSsystemwillnotprovidecost-effectivesecuritybyusing thesemethods.Therefore,thenecessityofintroducingamore accessibletechniquewithhigheraccuracyisclearlyobservable.
3. Waveletbasedanti-spoofingapproach
This paper presents an anti-deception technique which is directlyexecutableafteruncoveringtheattack.Theconsidered signalparameters includethe position solution.Weused dif- ferenceinthelastauthenticandcurrentfakesignalstoextract spooferror.RecentresearchesrevealthatWTisanefficienttool fortheisolationandseparationofsignalsfromnoise(Xuan&
Rizos,1997).BypassingpositionresidualtothestatisticWTat thefirststepandde-noisinginthenextstep,positiondeviations duetotheattackcanbeextracted.Then,theestimatedposition ofthereceivedsignalhasbeencorrected.Thefollowingsubsec- tionofferssomeinsightintothemathematicalbasisofWTand itsapplicationinGPSsignalprocessingtobetterperceivethe suggestedmethod.
3.1. WavelettransformutilizationinGPSsignalprocessing WTisaspecificsignalprocessingtooltoextractinformation fromsignals.AstandardWTcanbeexpressedas:
w(s,τ)=x,ψs,τ= 1
√s
∞
−∞x(t)ψ∗
t−τ s
dt (1)
where*denotescomplexconjugation, wisthewaveletfunc- tion,sisthedilationfactor,τisthetranslation,ψs,τ(t)ismother wavelet(Xuan&Rizos,1997).Afterapplyinginitialsignalsto aWT,yieldedwaveletcoefficientscanbemanipulatedinmany
Stationary GPS Receiver
Input IF Signal
Counterfiet Position Navigation
Positioning Tracking Acquisition
Authentic Position
Reconstructing coefficients
Tresholding Spoof Authentic
Position Decomposing
SWT
Spoof &
Noise
- -
+ ∑ +∑
Fig.1.Proposedinterferencemitigationalgorithm.
ways to achieve certain results such as filtering, de-noising, sub-bandcoding, compressing,feature detecting,etc.(Xiang, Liao,Zeng,&Wang,2013).Firstly,CollinandWarnant(1995) introducedtheWTforthepurposeofGPScycleslipcorrection.
Ogaja,Rizos,Wang,andBrownjohn(2001)introducedtheWT toanalyzetheGPSresultsinastructuralmonitoringapplication.
InRef.(Azarbad&Mosavi,2014)afterapplyingSWTtothe doubledifferenceresiduals,themultipatherrorisextractedby de-noisinganditusedtocorrectpositionerrors.De-noisingof signalsusingtheWTisoursubjectinthisarticle.Muchinfor- mationaboutsignalsinmanyapplicationsusuallyliesinafew numbers ofwaveletcoefficientsthat possesslargeramplitude incomparisonwithothercoefficients. An appropriatethresh- oldcanputoutnoiseof asignal. Mainly,itisnotpossibleto completelyfilteroutthenoisewithoutharmingtheinitialsig- nal.However,the performancecanbeoptimizedbyselecting theproperWT.
3.2. Wavelettransformselection
The continuous WT (CWT) was developed as an alterna- tive approach to the short time Fourier transform (STFT) to overcometheresolutionproblem.Itisapparentthatanalytical equationsandintegralscanperformneither theSTFT,northe CWT.Therefore,itisnecessarytodiscretizethetransforms.This conductsWTtodiscretewavelets.Inmanypracticalapplications suchasGPS,thesignalofinterestissampled.Asimplesolu- tionis implementingthewaveletfilter banksfor construction of the multi-resolution time-frequency plane. Thisexpansion iscalleddiscreteWT (DWT)whichprovidessufficient infor- mationbothfor analysis andsynthesisof the original signal.
Through asignificant reductioninthe computationtime,it is considerablyeasiertoimplementwhencomparedtotheCWT (Xuan&Rizos,1997).SWTissimilartoDWTexceptthatthe signalisneversub-sampledandinstead,thefiltersareupsam- pledateachlevelofdecomposition.ByusingSWTinsteadof standardWT, created coefficients are morethan sufficient to reconstructtheoriginalsignal.Thiscausesadditionalchoicesto beselectedtherequiredcoefficientsandthentheperformanceof thealgorithmcanbeimprovedbyselectingbetteronesamong allexisting coefficients(Jumah,2013). Indeed,it canstretch, shiftandprocesseachscalesignalpreciselyandimprovesthe
abilityofWTsignalprocessing.Asaresult,thispaperattempts touseSWTforde-noisingprocess.
Another aspectinWT selectionisits dimension(Harvala, 2012). Complex WT is a complex-valued extension to the standard DWT. It is a two-dimensional WT which provides multi-resolution, sparse representation, andusefulcharacteri- zation of the structure of a signal (Jalobeanu, Kingsbury,&
Zerubia,2001).Inourapplication,positionresidualsofallthree coordinatesareappliedsimultaneouslytotheone-dimensional WT.Useof twoor moredimensionalWTincreasesthecom- plexityofalgorithm,whileitisnotneededhere.Theremainder ofthissectionfocusesonthemainbodyofsuggestedtechnique.
3.3. Performingtheproposedspoof-reductionalgorithmin positionstage
Thissectiondescribesanovelsolutionthatdecreasesspoof- ing effecton the navigationpart of GPS receiver. When the receivedsignalisamixtureofbothdesiredsignalandspoofing whichcanbeproducedbyintentionalorunintentionalsources, an appropriate algorithm under certain circumstancescan be designedtoreducespoofingeffect.Inthiscondition,themain objectiveofinterferencecancelationistoestimatethetrouble- maker signalandthensubtractitfromtheinputsignalthatis acombinationoftheoriginalandinterferencesignals.Figure1 describes thesuggested algorithm indetailswhen a spoofing attack is recognized. The theoreticalfoundation of thisalgo- rithmisbasedonthefollowingdiscussion,itincludessixmain steps:
Step1:Immediateinterferencedetermination
Recognitionisthemostimportantpointbeforeexecutingthe mitigationalgorithm.Thedetectordeterminesthestarttimeof mitigationprocedure.Thelastauthenticpositionbeforedetec- tioncanbemodeledas:
PosA(t)=PosR(t)+N(t) (2)
PosA(t)containsthecorrectpositionPosRhaswidebandadditive noise, N isfunctionsintimet tobe sampled.Nincorporates allsourcesofun-spoofinginterferencesoverthechannel.GPS noiseandinterferencesignalstendtofallintothesamerange
offrequencies.Inthiscase,thepositionsolutionafteracquiring spoofingsignalwillbedefinedasfollows:
PosC(t)=PosA(t)+PosS(t)=PosR(t)+PosS(t)+N(t) (3) wherePosS indicates the counterfeit location coordinates. In similarscenarios,withoutareliablespoofdetector,anewposi- tionerrorcanbeproducedduetothespoof-reductionalgorithm.
Ofcourse,ourproposedmethodologydoesnotmakeposition errorsinthecaseofexecutingongenuineGPSdata,butitwill increasetheprocessingtime.
Step2:Subtractinglastauthenticpositionfromcurrentfake coordinatestoextract“Spoof&Noise”
Eq.(4)showsthedifferenceofcoordinatesbeforeandafter spoofingdetection.
PosSN(t)=PosC(t)− ˜PosR(t)= ˜PosS(t)+N(t) (4) where ˜P osS(t)isextractedpositionerrorand ˜P osR(t)isthelast positionsolutionbeforespoofdetectioninthefirstepochand estimatedauthentic positionfrom the previousepoch inlater epochs.
Step3:Decomposethe“Spoof&Noise”withSWT
PosSN(t)isappliedtoSWT,immediatelyafterspoofrecog- nition.Nowthede-noisingprocessattemptstomodifywavelet coefficientssothatnoiseisextractedfromtheoriginalsignal, wherethe unknownsignalPosS(t)mustberetrievedfromthe noisydataPosSN(t).N(t)isawhiteGaussiannoisewithanoise levelofσ2.ThereconstructionofPosSN(t)inthewaveletdomain meansthetranslationofallthewaveletcoefficientsof N(t)to zerovaluebyaproperthreshold.
AsmentionedinSection3.2,weusedone-dimensionalSWT in (X, Y, Z) three dimensional coordinates, because the sig- nalparametersinthreedimensionalcoordinatesareprocessed simultaneously.Nextimportantstepistochoosethebestmother wavelet,whichusuallydependsonapplication.Herewechoose theHaarwaveletasthemotherwavelet.Thiswaveletisasim- plestpossibilitywaveletandeasytoimplementillustratingthe desirablepropertiesofwaveletsingeneral.First,itcanbeper- formedinO(n)operations;second,itcapturesnotonlyanotion ofthefrequencycontentoftheinput,throughexaminingitatdif- ferentscales,butalsocontainstemporalcontent,i.e.,thetimes atwhichthesefrequenciesoccur(Donoho&Johnstone,1994).
Table1 liststhecorresponding coefficientsofthe up-sampled g[n]andh[n]forHaarmotherwaveletatlevel2.
Table1
HaarmotherwaveletcoefficientsforDWTandSWT.
Filter DWT Up-sampled(SWT)
High-pass(g[n]) g[0]=−1/√ 2 g[1]=−1/√
2
g[0]=−1/√ 2 g[1]=0 g[2]=1/√
2 g[3]=0 Low-passfilter(h[n]) h[0]=1/√
2 h[1]=1/√
2
h[0]=1/√ 2 h[1]=0 h[2]=1/√
2 h[3]=0
SWTofHaarwaveletwithscalingfunctioncanbeexpressed byEq.(5):
Ψ(x)=
⎧⎪
⎪⎪
⎪⎪
⎪⎨
⎪⎪
⎪⎪
⎪⎪
⎩
1; x ∈
1,1 2
−1; x ∈ 1
2,1
0; O.W.
, ϕ(x)=
1; x ∈ (0,1)
0; O.W.
(5)
Step4:Thresholdingtheoperationtochangethecoefficients obtainedfrompreviousstage
Thresholding is performed by determining the method of reformation coefficients and the noise modeling (Mosavi &
Azarbad, 2013).Mainly,thereare twotypesof hardandsoft thresholding. Hard thresholding zeros out small coefficients resulting in an efficient representation which is not proper forde-noising.Softthresholdingdecreasescoefficientsbythe thresholdvaluetoexceedthemforsoftening.Generally,thisis usedforde-noisingapplications(Jumah,2013).Hereweuseda ruleofshrinkagetomodifythecoefficientsinstep3asdeveloped byBorre,Akos,Bertelsen,Rinder,andJensen(2007).
Step5:Reconstructingthede-noisedSWTcoefficientsfrom step4toobtain“Spoof”
WTprovidespooferrorthroughreconstructingthede-noised SWTcoefficientsasfollows:
P os˜ S(t)=
l∈Z
A2lϕ(4t−l) (6)
whereA2isapproximatecoefficientofSWTatlevel2andϕis thescalingfunction.Itisworthemphasizingthatinabsenceof otherinterferences,spoofingerroristheonlyinaccuracyofthe positionestimations,afterthede-noisingprocedureisfinished.
Asmentionedintheprevioussubsection,theSWTisaninherent redundantscheme,aseachsetofcoefficientscontainsthesame number of samples as the inputand therefore, for a decom- positionof Nlevels,thereisaredundancy of2N.Asaresult, thereconstructionprocedureisdifferentfromthestandardWT (Mosavi&Azarbad,2013).
Step6:Reducingtheextractedspooferrorfromtheprimary positiontoget“authenticposition”
Asexplainedabove,thenoiseoftheinputsignalwithspoof partisappliedtoSWT,andextractedbyde-noisingafterpass- ing of the above steps. Finally, the algorithm subtracts the extracteddisturbancefromthe primary fakeposition solution togetauthenticpositioninformationfortheGPSnavigationas:
P os˜ A(t)=PosC(t)− ˜PosS(t) (7) AsaprivilegefromFigure1,thisalgorithmmakesnomud- dlewhenappliedtolegacysignal.Inotherwords,ifthedetector announcesafalseattackalarmandthealgorithmisactivatedin normalmode,thefinalpositionwillbegenuineanyway.More- over,locationaccuracymaybemodifiedowingtothede-noising procedure.
Software GPS Receiver
GPS Signal Generator RF Signal Combiner GPS Antenna
Real-time GPS Front-end
Fig.2.Toplevelmodelforimplementedsystem.
4. Resultsandanalysis
Inordertoinvestigatetheperformanceofthesuggestedtech- nique,wehaveimplementedandtestedourproposedplanon batchlaboratoryandmeasurementdatasetsseparately(Baziar etal., 2015).Laboratory platform schemeof thetotalsystem isdemonstratedinFigure2.Thesystemwasimplementedasa SDRinMatlab.AllprocessingwasdoneonalaptopAcer5755G withi72.2GHzCPU.Thissectionwilldescribethedatacollec- tionprocessandthenprovideresultsofcompensating.During differenttests,wealmostcompensatespoofingerror,entirely.
4.1. Counterfeitdatageneration
The fraudulent data collectionprocedure provides abatch dataset.Inthisprocedure,twoparametersareeffective:delay timeandamplitudeofdelayedsignal.SincethepowerofGPS signalislowonthesurfaceoftheEarth,thepowerofthecon- structed signalcanbeincreased andadjusted higherthanthe authentictosuccessfullymisleadthetargetreceiverandprevent simpledetection(Baziaretal.,2015).Thefirstlaboratoryspoof- ingdatasetisproducedfromIFsignalfromthecollecteddataset.
We combined a SDR with a transmitting RF front-end for practical implementation of an intermediate attack. The processed signalin single-frequency GPS receivers takesthe form:
SL1CA=ACCi(t)Di(t)sin(wL1t+ϕL1) (8) whereAC isC/A code amplitude,Ci(t) isithPRN C/A code, Di(t)istheithPRNnavigationmessage,ωL1istheangularfre- quencyofL1signalandϕL1isL1signalphase.Theconstructed counterfeitsignalcanbewrittenas:
CC/A(t)=AACCAi (t)DAi (t)sin(wL1t+ϕL1A )
+ADCCDi (t)DDi (t)sin(wL1t+ϕDL1) (9)
whereAandDpresenttheauthenticanddelayedsignals,respec- tively. CC/A(t) is indeed spreading signal for deception. For generating this signal, we need to take the legacy signal as delayedone.Afterprovidingthefakedsignalandtransmitting, the receivedsignaltothe victimreceivercanbeexpressedas Eq.(10)andmodifiedasEq.(11)(Baziaretal.,2015):
RL1C/A(t)=SL1C/A(t)+CC/A(t) (10)
RC/A(t)≈CC/A(t) (11)
Intheseconddataset,wetriedtoreleasefromquantization errorduetotheA/Dinthefront-endmodule.Forthispurpose, we decided tocombine the RF signalsinstead of IF signals.
With regardtoourlaboratory equipment,thiscanbe feasible onlywithaGPSsignalsimulator.Inthisway,thedelayedsignal describedabovewasemergedfromasimulator.Inthisscenario, itisgenerallyassumedthatsimulator’soutputismuchthesame signaldirectlytakenfromtheGPSantenna.Thecorruptedsignal inthisattackcanbeexpressedas:
d(n)=S(n)+αS(n−τ) (12)
whereαisamplificationfactorandequalsto2hereandτisthe delayofcounterfeitsignal.Accordingtotheabove-mentioned modeling, αS(n−τ)isactually consideredasinterference ele- ment.Afterthe RF inputsignalisconverted intoadigital IF signalandbeforesatelliteacquisition,spoofingattackisapplied tothedata.Inthisway,threecategoryspoofingdataareyielded.
Table 2 lists some examples of each group andspecifies the positionerrorinEast,NorthandUp(ENU)coordinates.
WhereRMSrefertoposition’sdifferencebetweennaviga- tional solutions based on authenticand spoof signals, ΔUis heightdifferenceandΔEandΔNarevariedinsurfacehorizons.
Low position error includes positioning error less than 300 meters. Spoofingbetween300 and500mcalledintermediate spoofing. Errors higher than 500m are considered as high positionerror.Throughanalyzingtheeffectofspoofinginthe
Table2
Detailsofspoofingdatasets.
Spoofingdatasets Laboratory Measurement
E[m] N[m] U[m] RMS[m] E[m] N[m] U[m] RMS[m]
Lowpositionerror
13 39 57 70 9 73 81 109
12 53 110 123 16 47 103 114
8 44 213 218 28 48 104 118
Middlepositionerror
9 76 343 351 26 112 87 144
10 88 392 402 6 103 166 195
133 153 473 515 18 125 199 236
Highpositionerror
210 237 745 810 59 310 265 412
233 264 829 901 324 266 242 484
309 341 1092 1185 314 204 382 535
receiversoftware,we discoveredthat spoofingattacks mainly distortvisiblesatellites,pseudo-rangesandsatellites’position.
Sincethe navigationsolution andGPS coordinates arebased onthesespecifications,noticeablespoofingerrorsareseen.The remainderof thissection willanalyze acquired resultsof the algorithm.
4.2. Testresultsoflaboratoryspoofingdata
Theproposedalgorithmwasappliedtolaboratoryspoofing datasetandthenresultswerestored.Theefficiencyofthesug- gestedtechniqueimplementationinmitigationofpositionerror isillustratedinTable3.EachrowofTablerelatestotheresults oftheproposedanti-spoofingalgorithmbyaspecifiedmother waveleton different spoofing data set. The “mitigationaver- age”columnindicatesmeanofreductionpercentageforevery WT.Thedifferencebetweenthehighestandthelowestreduc- tionpercentageofdifferentspoofingdataforeachWTreported as “Tolerance”.Asit canbe seenfromTable3,Haar WTat level2hasthebestoperationinaveragefordifferentdatasets, whichmitigatesinterferencewithinanaverageof99.56%anda toleranceof3%.
To verify the performanceof the proposed algorithm, the operation ofarandomly selected datasetisevaluatedbelow.
Coordinatevariationsbeforeandafter applyingthesuggested algorithmaredepictedinFigure3.Asobservedfromthefigure, inthecaseof quickdetection,theproposedinterference can- celingtechniquenullifiespowerfullytheundesirabledeviation causedbytheattack.
4.3. Testresultsofmeasurementspoofingdata
Table4reportstheresultsoftheinterferencerejectionalgo- rithmonexperimentaldatasets.Similartolaboratorydatasets, thebestresultscanbeselectedeasilyinTable4.Itisevident from Table4 that theHaar WT that completelycompensates spoofingerrorofalldatasets.
Thissectionalsopresentssometestscenariosthathavebeen usedforevaluatingtheperformanceofthesuggestedalgorithm onmeasurementdataset.Figures4–6plotENUcoordinatesofa sampleexampleofundertestdatasetbeforeandaftertheattack.
Itisclearfromthefigures,initial70epochsbelongtoauthentic data.At epoch71, the anti-spoofingalgorithm is activeafter intermediateattackdetection.
In orderto noiseobviation, authenticposition coordinates inthefirst 70epochs areaveraged andused asgenuine posi- tion in anti-spoofing algorithm, spatially if there is not an intermediate detection mechanism. In this way,not only the abnormal deviations of position coordinates due to spoofing have been compensated, but also wavelet de-noising process smoothesauthenticpositionobservationstomoreincreasenav- igationaccuracy.Finally,wecouldexpressawhollyinterference reductiontechnique.
4.4. Performancecomparison
Becauseofdissimilaritybetweenofproposeddetectionand mitigation methods with existing ones, accurate comparison
Table3
Detailsofinterferencerejectionalgorithmperformanceonlaboratoryspoofingdata.
WT Level Positionerror[m] Mitigation
average(%)
Tolerance(%)
70 123 217 351 402 514 810 900 1185
Mitigationpercent
Haar 2 97 99 100 100 100 100 100 100 100 99.56 3
Dmey 3 91 94 99 99 99 100 100 100 100 98 9
sym-12 3 93 96 99 99 99 100 100 100 100 98.44 7
db-10 4 91 95 99 99 100 100 100 100 100 98.22 9
coif-5 3 91 96 99 99 100 100 100 100 100 98.33 9
bior-6.8 4 91 96 100 100 100 100 100 100 100 98.56 9
rbio-5.5 3 91 95 99 100 99 100 100 100 100 98.22 9
10
10
10
20
Positioning Epochs
Positioning Epochs
Positioning Epochs
30 40
a
b
c
60
60
60 70
70
70 50
50
50 40
40 30
30 20
20 2000
2000
5000 -2000
-5000 -2000 0
East Error [m]North Error [m]Up Error [m]
0
0
0
0
0
Spoofed system Modified system
Fig.3.ENUcoordinatesbeforeandafterapplyingtheinterferencerejectionalgorithm.
Table4
Detailsofinterferencerejectionalgorithmperformanceonexperimentaldatasets.
WT(level) Positionerror[m] Mitigation
average(%)
Tolerance(%)
109 114 118 144 195 236 412 484 535
Mitigationpercent
Haar(2) 100 100 100 100 100 100 100 100 100 100 0
Dmey(3) 93 94 94 97 94 93 98 100 99 95.78 7
sym-12(3) 80 82 84 99 98 99 100 98 99 93.22 20
db-10(4) 89 89 92 98 98 99 100 100 99 96 11
coif-5(3) 99 97 98 99 100 100 100 100 100 99.22 3
bior-6.8(4) 100 98 100 99 99 100 100 100 100 99.56 2
rbio-5.5(3) 88 83 89 99 99 100 100 100 100 95.33 17
withprior works isdifficult.Table 5 producesacomparative evaluationof newandprevious mitigation methodsbased on complexity,necessaryequipment,advantagesanddisadvantages (Jafarnia-Jahromi,Broumandan,Nielsen,&Lachapelle,2012).
Forreliableandcorrectjudgment,weassignedanumerical valuetoanyfeature.Theworstandbestcaseisconsideredfor
anyfuture;0scoreisdedicatedforworststateand5scoreforthe best.Then,dependingonalgorithmoperationanumberfrom0 to5isassignedtoanyfeature.Forexample,about“necessary equipment”feature, analgorithmearns5 scoresif neededno extraequipment,andinthecaseofnecessitytobasalchanges inreceiverstructure,ittakes0scores.Resultsofnumberingare
5.35
5.35 5.36
5.36 5.365
5.365 5.37
5.37 5.375
5.375 5.355
5.355
20
East Coordinate (m)East Coordinate (m)
20 0
0 x 105
x 105
40 60
a
b
60 40
80
80 100
100 120 140
140 120
Positioning Epochs
Positioning Epochs
Fig.4.EastENUcoordinates:(a)beforeand(b)afterapplyingtheproposedalgorithm.
0
0
20
20 North Coordinate (m)North Coordinate (m) 40
40
60
60
80
80
100
100
120
120
140
140 x 106
x 106 3.954
3.954 3.952
3.952 3.9525
3.9525 3.953
3.953 3.9535
3.9535
a
b
Positioning Epochs
Positioning Epochs
Fig.5.NorthENUcoordinates:(a)beforeand(b)afterapplyingtheproposedalgorithm.
0
Up Coordinate (m)Up Coordinate (m)
5000
5000 -5000
-5000 0
0 10000
10000
0
20
20
40
40
60
60
80
80
100
100
120
120
140
140 Positioning Epochs
Positioning Epochs
a
b
Fig.6.UpENUcoordinate:(a)beforeand(b)afterapplyingtheproposedalgorithm.
illustratedinFigure7.Asitcanbeseen,theproposedalgorithm gets12pointsaffirmingitisbetterthanothers.Signalestimation (Linetal.,2007)andVB(Jafarnia-Jahromi,Lin,etal.,2012) methodsareimplementedintrackingloop.Thefirstoneisnot effectiveforlongattacksandthesecondonemakesamassive changeinthereceivertrackingloopstructure.Spatialprocessing (Broumandanetal.,2015)isaneffectiveandreliabletechnique, butbecauseof addingextrahardware,havehighimplementa- tioncosts,too(Magiera&Katulski,2015).NewRAIMmethods
(Ledvinaetal.,2010)arebeingdeveloped,butthesealgorithms arecomplexandmaybedifficulttoimplementrobustly.Ifsuch algorithmstobesucceed,typicallytheymustachievedetection atthemomentofsignaldrag-off,whichdegradestheirrobust- ness.Asitcanbeseen,theoveralleffectivenessoftheproposed methodissuperiortoothers.Inadditiontothebenefitsof the previousmethods,suggestedtechniqueinthispaperrequiresno additionalhardwareandhassimplerimplementation,yetitisan accuratemethod.
Table5
Comparativeperformanceofspoofmitigationtechniques.
Mitigationtechniques Algorithmlocation Necessaryequipment Complexity Advantages Disadvantages Signalestimation(Linetal.,2007) Navigation Inertialsensoror
estimator
Low Simplicity Errorgrowthinlongattacks Spatialprocessing(Broumandanetal.,2015) IFsignal Antennaarray Medium Noneedforprevious
information
Highcostandinefficiencyin multi-antennaspoofer VB(Jafarnia-Jahromi,Lin,etal.,2012) Trackingloop Extratrackingloop High Highaccuracy Highimplementationcosts RAIM(Ledvinaetal.,2010) Navigation Softwareupgrading Medium Highaccuracy Unreliable
Thiswork Navigation Softwareupgrading Low Real-time,reliable
andhighaccuracy
Needtopastinformation
This work
VB
Spatial Processing
Signal estimation
RAIM
0
Complexity Disadvantages Advantages Equipment Total
2 4 6 8 10 12 14 16
Fig.7.Performancecomparisonofspoofingmitigationalgorithms.
Table6
Timecomplexityofproposedmethod.
Spoofingdatasets RMS [m]
SDR execution time[s]
Addedtime dueto algorithm[s]
Percentof added time(%)
Lowpositionerror
109 281.677 4.298 1.53
114 311.231 3.977 1.28
118 311.121 4.161 1.34
Middlepositionerror
144 159.163 5.122 3.22
195 283.446 4.374 1.54
236 331.144 4.273 1.29
Highpositionerror
412 292.983 3.889 1.33
484 328.812 3.599 1.09
535 320.123 4.188 1.31
Moreover,unliketheothertechniques,successofthisalgo- rithmdoesnotdependonthekindofattack.Timecomplexityof implementedalgorithminMatlabsoftwareisshowninTable6.
Asitcanbeobserved,executiontimediffersforvariousdatasets.
Thisisduetodifferenceofline-of-sitesatellitesforeachdata set.Obviously,manysatellitestakelongertimetoexecute.For example,spoofdatawith109mpositionerrortakes281.677s.
After implementing,execution time increases as 4.298s that equates with 1.53% growth. In this way, it is inferable that theoveralleffectivenessoftheproposedmethodissuperiorto others.Eventhough,itneedsmodificationatnavigationstage.
5. Concludingremarks
Thispapercontainsareviewofliteraturepertainingtospoof- ing attacks, interference rejection techniques and WT with specificinterestintheirstrongpointandproblems.Moreover,a newinterferencemitigationtechniqueusingwaveletde-noising hasbeenproposed.Twotypesofcounterfeitdatasetareused forevaluatingtheproposedalgorithm.Designingthisapproach has been repeatedly tested and its performance was demon- stratedintermsofpositionerrors.Itcanbeseenfromtheresults thattheRMSvaluesofthespoofingerrorshavebeenreduced
fullyafterapplyingthecorrections,whichlooksonasoutstand- ingperformanceininterferencesuppressionfield.Finally,asa greatpleasure,thecivilstationaryGPSreceiverscanbeimme- diatelymodifiedtoexploittheproposedauthenticationstrategy.
Moreover,itisexpectedthatbecauseofexecutingonposition solutiontheproposedtechniquecanbeeffectiveinotherkinds ofinterference,whichwillbeinvestigatedinfutureworks.
Conflictofinterest
Theauthorshavenoconflictsofinteresttodeclare.
References
Azarbad,M.R.,&Mosavi,M.R.(2014).Anewmethodtomitigatemulti- patherrorinsingle-frequencyGPSreceiverwithwavelettransform.GPS Solutions,18(2),189–198.
Baziar, A.R., Moazedi,M., & Mosavi,M.R. (2015).Analysis ofsingle frequencyGPSreceiverunderdelayand combiningspoofingalgorithm.
WirelessPersonalCommunication,83(3),1955–1970.
Borre,K.,Akos,D.M.,Bertelsen,N.,Rinder,P.,&Jensen,S.H.(2007).A software-definedGPSandGalileoreceiver:Asingle-frequencyapproach.
Birkhäuser.
Broumandan, A.,Jafarnia-Jahromi, A.,& Lachapelle, G.(2015). Spoofing detection,classificationandcancelation(SDCC)receiverarchitecturefor amovingGNSSreceiver.GPSSolutions,19(3),475–487.
Collin,F.,&Warnant, R.(1995).Applicationofthewavelettransformfor GPScycleslipcorrectionandcomparisonwithKalmanfilter.Manuscripta Geodaetica,20,161–172.
Donoho,D.L.,& Johnstone,I.M.(1994).Thresholdselectionforwavelet shrinkage of noisydata. pp.A24–A25. IEEEEngineering in Medicine andBiologySociety,1994.Engineeringadvances:Newopportunitiesfor biomedicalengineers.Proceedingsofthe16thannualinternationalconfer- enceoftheIEEE(Vol.1).
Harvala,D.N.(2012).Alowpowerapplication-specificintegratedcircuitimple- mentationofwavelettransform/inversetransform(1sted.).BiblioScholar.
Jafarnia-Jahromi, A., Broumandan, A., Nielsen, J., & Lachapelle, G.
(2012). GPS vulnerability to spoofing threats and a review of anti- spoofingtechniques.InternationalJournalofNavigationandObservation, 1–16.
Jafarnia-Jahromi,A.,Lin,T.,Broumandan,A.,Nielsen,J.,&Lachapelle,G.
(2012). Detection andmitigation ofspoofing attackson avector-based
trackingGPSreceiver.InternationalTechnicalMeetingoftheInstituteof Navigation,1,790–800.
Jalobeanu,A.,Kingsbury,N.,&Zerubia,J.(2001).Imagedeconvolutionusing hiddenMarkovtreemodelingofcomplexwaveletpackets.InIEEEinter- nationalconferenceonimageprocessing(pp.201–204).
Jumah,A.I.(2013).Denoisingofanimageusingdiscretestationarywavelet transformandvariousthresholdingtechniques.JournalofSignalandInfor- mationProcessing,4(1),33–41.
Ledvina,B.M.,Bencze,W.J.,Galusha,B.,&Miller,I.(2010).Anin-lineanti- spoofingdeviceforlegacycivilGPSreceivers.InInternationaltechnical meetingoftheinstituteofnavigation(pp.689–712).
Lee,J.H.,Kwon, K.C.,An,D.S.,&Shim,D.S.(2015).GPSspoofing detectionusingaccelerometersandperformanceanalysiswithprobabilityof detection.InternationalJournalofControl,AutomationandSystems,13(4), 951–959.
Lin,Z.,Haibin,C.,&Naitong,Z.(2007).Anti-spoofingextendedKalmanfil- terforsatellitenavigationreceiver.InInternationalconferenceonwireless communications,networkingandmobilecomputing(pp.996–999).
Magiera,J.,&Katulski,R.(2015).DetectionandmitigationofGPSspoof- ingbasedonantennaarrayprocessing.JournalofAppliedResearchand Technology,13(1),45–57.
Mosavi,M.R.,&Azarbad,M.R.(2013).Multipatherrormitigationbasedon wavelettransforminL1GPSreceiversforkinematicapplications.AEU– InternationalJournalofElectronicsandCommunications,67(10),875–884.
Mosavi,M.R.,Nasrpooya,Z.,&Moazedi,M.(2016).Advancedanti-spoofing methodsintrackingloop.JournalofNavigation,69(4),883–904.
Ogaja,C.,Rizos,C.,Wang,J.,&Brownjohn,J.(2001).Towardtheimple- mentationofon-linestructuralmonitoringusingRTK-GPSandanalysisof resultsusingthewavelettransform.In10thinternationalsymposiumon deformationmeasurements(pp.284–293).
Xiang,F.,Liao,G.,Zeng,C.,&Wang,W.(2013).Amultipathmitigationdis- criminatorforGPSreceiver.AEU–InternationalJournalofElectronicsand Communications,67(10),839–847.
Xuan,F.W.,&Rizos,C.(1997).TheapplicationsofwaveletstoGPSsignal processing.In10thinternationaltechnicalmeetingofthesatellitedivision oftheinstituteofnavigation(pp.697–703).
