649
1 INTRODUCTION
In times when satellite navigation literary lies in
foundations of modern society (UK Government
Office of Science, 2018), every threat to GNSS
Positoning, Navigation, and Timing (PNT) services
should be considered with due attention. GNSS
spoofingisamaliciousactionsaimedatdegradation
of GNSS PNT services, with
potential huge
consequencesonlife,safety,security,health,property
and economy. Identified as an information security‐
relatedissue,GNSSspoofingisthetargetofgrowing
numberofresearchstudiesacrosstheworld.
Here a novel concept of GNSS spoofing counter‐
measureisproposed,establishedonidentificationof
GNSSpositionestimation
processandGNSSreceiver
design shortcomings and vulnerabilities,
characterisationofthenatureofGNSSspoofing,and
survey of recent research in GNSS anti‐spoofing
methodsdevelopment.TheGNSSSpoofingDetection
andMitigation(GNSSSDM)methodisfoundedona
dedicated architecture and capabilities of modern
computational and communication technologies. It
does
notrequirecoreGNSSsystemmodification,and
allows for seamless transition of majority of GNSS‐
enableddevicestowardssafetyfromGNSSspoofing.
The manuscript reads as follows. This Section
outlines the motivation for research, and introduces
the content of the manuscript. Section 2 details the
GNSSposition estimation mathematical
method and
procedure,whileidentifyingpotentialvulnerabilities.
Section 3 addresses recent trends in GNSS receiver
design, especially a more direct implementation of
mathematical methods using software solutions,
rather with model approximation using electronic
circuitry.Section4 outlinestheresultsoftheprevious
Foundations of GNSS Spoofing Detection and
Mitigation with Distributed GNSS SDR Receiver
M.Filić
UniversityofLjubljana,Ljubljana,Slovenia
ABSTRACT:GNSSspoofingisanintentionalandmaliciousactionaimedatdegradingandsuppressingGNSS
Positioning,Navigation,andTiming(PNT)services.SinceitaffectsdataandinformationsegmentofGNSS,itis
consideredaGNSSinformation(cyber‐)securityattack.Consideringasignificantandpowerfulthreat,
GNSS
spoofingshouldbetreatedseriouslytoavoiddamageandliabilitiesresultingfromdisruptionsofGNSSPNT
services.HeretheGNSSpositionestimationprocedureisexaminedforpotentialvulnerabilities,andthenature
of and motivation for GNSS spoofingattacks exloiting the vulnerabilities assessed. A novel GNSS Spoofing
Detection and Mitigation
(GNSS SDM) method is proposed within the established computational and
communication infrastructure, that allows for successful overcoming and classification of GNSS spoofing
attacks.ProposedmethodisapplicablewithoutrequirementsforcoreGNSSmodification,andleavesmajority
ofuserequipmenteasilytransferabletotheGNSSspoofing‐freeenvironment.PotentialGNSSspoofingeffects
andGNSSanti‐spoofingopportunitiesinmaritimesectorweregivenaparticularattention.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 12
Number 4
December 2018
DOI:10.12716/1001.12.04.01
650
researchsurvey,andsummarisespotential approach
in GNSS spoofing counter‐measures development.
Section5 presentsthe detailsof the proposed GNSS
SDM method, the architecture in support of GNSS
SDM method implementation, results of practical
validation and concluding comments. Section 6
addresses potential benefits of the GNSS SDM
implementation in
maritime sector. Manuscript
concludeswithsummariesoffindingsandproposals
for further research in Section 7, and with an
exhaustivelistofreferences.
2 GNSSPOSITIONESTIMATIONMODEL
Thesatellite‐basedpositionestimationreliesuponthe
accuratemeasurementof satellitesignalpropagation
time between satellite and receiver aerials, and on
provision of accurate position of every satellite
involvedinestimation process (PetrovskiandTsujii,
2012), and (Filić and Filjar, 2018a) for discussion. A
GNSS receiver measures the satellite signal
propagation time (and, therefore, the distance,
pseudorange, between the satellite and receiver
aerials,providingthesatellitesignalpropagatesatthe
constant velocity
equal to the velocity of light in
vacuum) using a cross‐correlation‐based statistical
signal processing procedure (Petrovski and Tsujii,
2012) applied on the Pseudo‐Random Noise (PRN)
codesbroadcastbyGNSSsatellitesascomponentsof
the composite GNSS signals. PRN codes uniquely
identifiesaGNSSsatelliteandallowsforaccurate
and
precise propagation time measurement. (Petrovski
and Tsujii, 2012) reports research initiatives to
modernise GNSS systems and service through
replacement of currently used PRN codes with
chaotic signals. The transition may allow for more
accurate measurements through improved
synchronisation (Pecora et al, 1997), and enhanced
resilience against natural and artificial (Alvarez
and
Li, 2006) sources of interference, as it has been
demonstrated in theory and in the other
telecommunication systems (Sun, 2015), (Babu and
Singh, 2013).(Vaudenay, 2005), (Alvarez and Li,
2006)and(Sun,2015)discussedopprtunitiesofchaos‐
based cryptography in securing communication
systemsandadvancingthequalityoftheir
services.
Broadcast satellite signal may be formally
describedasin(1),givenisthedescriptionoftheUS‐
operated GPS composite signal structure for
commercial‐grade single‐frequency GPS receivers
(PetrovskiandTsujii,2012).
111
11
2
2
GPSL p i i L L
ci i L L
saPtDtsinft
aC t D t cos f t
(1)
where:
… denotes broadcast composite GPS satellite
signal transmitted on L1 = 1575.420 MHz carrier
(commonly used by commercial‐grade single‐
frequencyGPSreceivers)
i…indexofthesatelliteinconsideration(i‐th)
t…timeinstant
a
p… denotes amplitudeof authorisedcomponent of
broadcastcompositeGPSsatellitesignal
P
i(t)…denotesthePrecisionGPSPRNbinarycodefor
pseudorange measurement provided to authorised
(dual‐frequency)GPSusers
D
i(t) … denotes binary coded navigation message,
with satellite ephemeris data for satellite position
determination, parameters for error correction
models, satellite and system health status
information,andtheotherrelateddatasets
f
L1=1575.420MHz
φ
L1…phaseofL1carrier
a
c … denotes amplitude of the Coa r se Acquisition
(C/A)componentofbroadcastcompositeGPSsatellite
signal
C
i(t)… denotes the Coarse Acquisition (C/A) GPS
PRN binary code for pseudorange measurement
providedtoall(bothauthorised, dual‐frequency,and
civil,single‐frequency)GPSusers
A dedicated GNSS statistical signal processing
procedure within a GNSS receiver returns a
pseudorange estimate comprising the actual (true)
distance between a satellite and a
receiver aerials,
denoted with R, anda number of errorcomponents
that compromise measurements process, and cause
theGNSSpositionestimationerror.Thepseudorange
observationstructureisgivenwith(2)(FilićandFiljar,
2018a).
rec ionospheric tropospheric
multipath sat eph sat clock random
Rct
(2)
where:
ρ … denotes a pseudorange, measured using
statistical signal processing procedure in GNSS
receiver,betweenthesatelliteinobservationandthe
userGNSSreceiveraerialin[m]
c … denotes velocity of electromagnetic wave
propagationinvacuum
δt
rec…denotesuserGNSSreceiverclockerror(large,
unknown, but statistically independent from the
choiceofsatellite)
ε
ionospheric … denotes pseudorange measurement error
duetoionosphericeffects(FilićandFiljar,2018a)
ε
tropospheric … denotes pseudorange measurement error
duetotroposphericeffects(PetrovskiandTsujii,2012)
ε
multipath … denotes pseudorange measurement error
duetomultipatheffects(PetrovskiandTsujii,2012)
ε
sat‐eph … denotes pseudorange measurement error
duetosatelliteephemeriserrors(PetrovskiandTsujii,
2012)
ε
sat‐clock … denotes pseudorange measurement error
due to variations in satellite clockaccuracy
(PetrovskiandTsujii,2012)
ε
random … denotes the other pseudorange
measurementerrors,uncorrelatedtopreviousgroups
and of random nature (Filić, Filjar, 2018) (Petrovski
andTsujii,2012)
TheaimoftheGNSSpositionestimationprocessis
to yield the measurement‐based unambiguous
estimate of user GNSS aerial position in three‐
dimensional WGS‐84 coordinate
system, time in
Universal Time Co‐ordinated (UTC) system, and
GNSS positioning error vector. The GNSS‐based
description of user state comprises four spatio‐
temporalstatevariables,aspresentedwith(3),where
x,y,andzdenotesthethreecomponentsofa position
651
estimate vector in respective WGS‐84 datum frame
(PetrovskiandTsujii,2012).
,,,
rec
x
xyz t
(3)
The GNSS problem (3) solution is commonly
obtained using the iterative procedure given in (4)
and(5)(PetrovskiandTsujii,2012),(Filić,2017),(Filić,
Grubišić and Filjar, 2018), (Filić and Filjar, 2018a),
wherekdenotestheiterationstep.
1
1
1
kk
kk
kk
x
xx
y
yy
zzz
(4)
111
1, 1, 1,
11,
223
2, 2, 3,
22,
33,
333
3, 3, 3,
4
444
4, 4, 4,
kS kS kS
kkk
k
kS kS kS
kkk
k
k
kS kS kS
kkk
rec
kS kS kS
kkk
xx yy zz
c
RRR
R
xx yy zz
x
c
RRR
R
y
R
z
xx yy zz
c
RRR
ct
xx yy zz
c
RRR
4,k
R
(5)
where:
(x
k,yk,zk)…denotesuserGNSSpositionestimateink‐
thiterationstep
(x
Si,ySi,zSi)…denotesi‐thsatellitepositionatthetime
ofsatellitesignalbroadcast
{ρi, i = 1, …, 4} … a set of GNSS pseudorange
measurements taken with four different satellites
independentlyandatthesametime
222
,ik k Si k Si k Si
Rxxyyzz (6)
The 4 x 4 matrix in (7) is commonly known as
geometric(G)matrix.
111
1, 1, 1,
223
2, 2, 3,
333
3, 3, 3,
444
4, 4, 4,
kS k S kS
kkk
kS k S kS
kkk
kS k S kS
kkk
kS k S kS
kkk
xx yy zz
c
RRR
xx yy zz
c
RRR
G
xx yy zz
c
RRR
xx yy zz
c
RRR
(7)
System(4)and(5)maybedeployedfordirectuser
state (positionand time) estimation under condition
ofsuccessfulsuppressionofalltheerroreffects(Filić,
2017), (Filić, Grubišić, and Filjar, 2018), (Filić and
Filjar,2018a). However,if the residualerrors cannot
be neglected, an advanced optimisation approach
should be taken to assure the solution stability and
quality of position estimates (Filić, 2017), (Filić and
Filjar,2018a).
Weighted Least‐Square position estimation
methodmaybeconsideredthesolutionoftheabove‐
statedoptimisationproblem(Filić,2017),(Gustafsson,
2010),(GallierandQuaintance,2018).Ingeneral,the
Weighted
Least‐Square method aims at solving the
problem (8), minimising the square error between
observedandestimatedvalueofavariabley.
2
ˆ
iW
miny y
(8)
Consideringunequalcontribution(importance)of
particular predictors (input variables), the
minimisation problem solution may be expressed
using weights
w
, as expressed in (9) (Gustafsson,
2010).
2
2
1
ˆˆ
n
iW ii
i
miny y w y y
(9)
Weighted Least‐Square method yields the GNSS
navigation problem solution in the form of (10),
where
w
x
denotes weighted user state estimate, G
denotes geometric matrix, W denotes matrix of
weights,and
y
denotes vector ofGNSS pseudorange
measurements (Filić, 2017), (Filić and Filjar, 2018a).
Weightsmay be defined ina mannertoaddress the
particular effect of known statistical description, as
discussed and demonstrated in (Filić, 2017) for the
problemofuncorrectedrandomcomponentofGNSS
ionosphericdelay.
1
w
x
GWG GWy
(10)
Eq (10) yields a single point solution of GNSS
navigation problem, which may be corrupted still
with residual errors, mostly of stochastic nature.
FurtherclearingofGNSS‐basedpositionestimatesis
commonly performed through implementation of
Kalman filter for removal of Gaussian position
estimation errors (Filić and Filjar, 2018a),
(Petrovski
andTsujii,2012).
3 GNSSSOFTWARE‐DEFINEDRADIO(SDR)
RECEIVERCONCEPT
Position estimation process based on GNSS
observationstakesplaceinauserGNSSreceiver.This
servicearrangementallowsforprivacyprotectionand
the GNSS position estimates to remain in user
equipmentonly,sincenocommunicationresponseis
eversent
fromaGNSSreceivertoaGNSSsatellitein
acorepositioningservicescenario,underconditionof
the user’s consent. Information on the user’s
whereabouts may be exchanged for operation of
GNSS‐based applications,but this extends the scope
ofthecoreGNSSpositioningservice.
The GNSS position estimation procedure
comprises signal and data processing at the three
essentialdomains,asdepictedinFigure1.: (i)Radio
Frequency(RF)domain,(ii)Base‐Band (BB) domain,
and (iii) Navigation domain. Figure 1 outlines the
signals and data considered in related domains, as
well as results of the domain‐related processing.
Detailed description
may be found elsewhere
(PetrovskyandTsujii,2012),(FilićandFiljar,2018a).
652
A traditional GNSS receiver deploys an
electronics‐based approach in processing GNSS
signals and data, where dedicated and tailored
electronicscircuitry is utilisedtoperformprocessing
tasks governed by mathematical methods and
models. In recent years, the Software‐Defined Radio
(SDR) approach has gained popularity. Aimed at
replacement of every
possible electronics‐based
solutionwithasoftware‐basedsolutionoperatedona
general purpose hardware (such as a personal
computer, or a smartphone), SDR allows for more
accurate, efficient, flexible and re‐configurable
receiver architecture. Those are accomplished with
the direct implementation of mathematical methods
and models in software, instead
of approximating
them with electronics circuits (Stewart et al, 2015),
(FilićandFiljar,2018a).
Origins of the Software‐Defined Radio may be
traced to research conducted by Joseph Mitola III
(Mitola, 1995). A comprehensible and problem‐
oriented coverage of the subject may be found in
(Stewartetal,2015).(Petrovskiand
Tsujii,2012)and
(Filić and Filjar, 2018a) gave a systematic view on
SDR utilisation in satellite navigation, including a
briefoutlinegiveninthisSection.
GNSSisatelecommunicationsystem,andaGNSS
receiver is its component. With a GNSS receiver
complying to common procedures of a
telecommunication system, it
has been a
straightforwardactiontorenderatransitionfromthe
traditionaltoSDR‐basedGNSSSDRreceiverdesign.
GNSSSDR receiverhas advancedin themeans that
the Base‐Band (BB) and Navigation domain signal
anddataprocessingmovedfrom(mostly)hardware‐
based to (entirely) software‐based. Such a transition
has opened a wide perspective for advanced and
independent development of error correction and
position estimation methods and models, tailoredto
satisfy targeted requirements of GNSS‐based
applications (systems and services) (Filić and Filjar,
2018a).
Figure1.GNSSsignalandinformationprocessinginGNSS
Software‐DefinedRadio(SDR)receiver
Avastroomforopportunitieshasbeenopenedin
both the BB and Navigation domains for
improvements and enhancements. Additionally,
recent developments in gaining access to raw GNSS
observations (GNSS pseudoranges and navigation
message data) in smartphones through transparent
interface between BB and Navigation domains (Filić
and Filjar, 2018b) attracts research
interests in
exploiting new source of massive data sets bearing
marks of positioning environment effects, both
natural (ionospheric and multipath effects) and
artificial(spoofingandjamming).
The advancement prospectshave arisen from the
methodology of data processing in Navigation
domain,asdepictedinFigure2.Theimplementation
of a selected
GNSS position estimation method is
preceded by raw GNSS observations, commonly
pseudoranges,assessmentsandcorrectionsforknown
sourcesofsystematicerrors.Ionosphericdelayeffects
that corrupt pseudorange measurements are
commonly corrected using NeQuick (for Galileo
pseudoranges) or Klobuchar (for GPS and Bediou
pseudoranges) error correction models, which
parameters are sent as
a part of navigation message
(Filić and Filjar, 2018) (Petrovski and Tsujii, 2012).
Tropospheric effects on GNSS pseudoranges are
corrected using Saastamoinen or Neill mapping
functionerrorcorrectionmodels(PetrovskiandTsujii,
2012). Satellite clock errors are corrected using
stochastic models with parameters broadcast in
navigationmessage,again.
GNSSpseudoranges
correctedforforeseeableand
known systematic errors are used for GNSS‐based
positionandpositioningerrorestimation,asoutlined
inSection2ofthismansucript.
Considering recent developments in computer
science, telecommunications, and signal processing
disciplines,theGNSSSDRconceptmaybeenhanced
further with the proposal of distributed GNSS SDR
receiver concept. With the reference to Figure 3, a
novel GNSS SDR receievr design is proposed
herewith, that leaves the RF component only in a
mobile part of a GNSS SDR receiver, together with
provision of reliable and uninterrupted
communicationwiththerestof the receiver (BBand
Navigationdomain
processors),hostedbyanexternal
facility,inthecomputingenvironmentwithasuitable
computational capacity and the access to position
estimation assistance data (for instance, tailored
local/regional error correction model, bespoke
position estimation method that complies to
requirementsoftargetedGNSSapplicationetc.).
Figure2. GNSS information processing in Navigation
domain
In implementation of a distributed GNSS SDR
receiver,mobileunitwouldsendaseriesofsnapshots
of digitised waveforms of demodulated composite
GNSSsignalcomponents(C/APRNcodes,navigation
message data/signals) to the BB and Navigation
domain processing facility (running in cloud, as an
example) for the complete observations‐based
position
estimation, thus allowing for personalised,
653
assisted and more accurate position estimation, that
will spare mobile unit from power‐consuming
extensivecomputationwithinsufficientassistance.
4 GNSSSPOOFING
GNSS spoofing is an intentional malicious and
pernicious information security attack on GNSS,
conducted through broadcast of counterfeit
(manipulated)GNSSsignalswithmaliciousintention
of deception of GNSS
receiver to estimate incorrect
positionusingallegedlyoriginalGNSSsignals.
With GNSS becoming an essential component of
national infr astructure, resilience against any GNSS
Positioning, Navigation, and Timing (PNT) service
distortionhasbecometheprioritisedtasktoresolveto
allow for seamless and uninterrupted operation of
GNSS‐based applications. Numerous often
multi‐
disciplinary research groups affiliated to respected
researchorganisationaddressedtheproblemintheir
research studies.(Schenewerk et al, 2016) outlined a
procedure for navigation message examination for
quality assurance in the matters of archiving the
material with the International GNSS Service (IGS)
repository (Jafarnija‐Jahromi et al, 2012) presented a
detailed
studyoftheproblem,itsnature,technology
background, and possible approaches for counter‐
measures development. (Tippenhauer et al, 2011)
presented a thorough study of requirements to be
fulfilledforasuccessfulGNSSspoofingcyber‐attack.
T E Humphreys and his team defined the problem
first, and then examined a number
of different
scenariosofGNSSspoofingattacks,proposingseveral
approaches to GNSS anti‐spoofing methods
development in (Humphreys et al, 2008),
(Nighswander et al, 2012), (Wesson, Rothlisberger,
andHumphreys,2012),(Kernsetal,2014a),(Kernset
al, 2014b), and (Psiaki and Humphreys, 2016). The
group even demonstrated successful GNSS spoofing
attacks
that resulted with remote overtaking control
of a vessel (Bhatti and Humphreys, 2017) and an
aircraft(Kernsetal,2014a).
Survey of research accomplishments so far,
conducted during the study presented here, reveals
the Base‐Band (BB) domain as the target of GNSS
spoofing methods. Attacks focus generally on both
PRN signals and navigation message data forgery,
thuspersuadingaGNSSreceivertodetermineapre‐
engineeredincorrectposition.Still,majorityofGNSS
spoofing attacks address modification of GNSS
navigationmessageasamoreefficientandpragmatic
meansforGNSSPNTservicesdistortion.
TheGNSSspoofingcounter‐measures(GNSSanti‐
spoofing)
concentrateonpatchingtheinheritedGNSS
vulnerabilities in all three domains of GNSS signal
and data processing (Figure 1). GNSS anti‐spoofing
methods rely almost entirely on modification of the
coreGNSSsystem,withtheonlyexceptionofmulti‐
aerialuserequipmentutilisation.(Humphreysetal,
2008) identified the following
candidate GNSS anti‐
spoofing measures: (i) amplitude discrimination, (ii)
time‐of‐arrival discrimination, (iii) assessment of
consistency of navigation inertial measurement unit
(IMU), (iv) discrimination of polarisation, (v) angle‐
of‐arrival discrimination, and (vi) cryptographic
authentication. All the proposed approaches assures
user equipment independence from any kind of
communicationnetwork.Assurance
thatthereceived
signalandnavigationmessagearegenuine(original)
is the aim of the GNSS anti‐spoofing methods. The
matter has recently become known as the GNSS
signal authentication requirement (O’Driscoll, 2018),
(Capparra et al, 2016). Its development and
implementationisincreasinglyregardedtoutilisation
of tailored encryption (Wesson,
Rothlisberger, and
Humphreys, 2012), (Kerns et al, 2014b) that will
provideauthenticityassurance,andevenenhancethe
GNSSpseudorangemeasurementaccuracy,unrelated
toitstheinitialpurpose.Aprospectofutilisationofa
‘trusted’ receiver for GNSS spoofing detection was
examined in (Kuhn, 2010). The concept may be
extended with consideration
of scenarios such as
EGNOSassistancedatastreamingservice,orA‐GNSS
service provided by mobile telecommunications
networks as the means of provision of GNSS data
from ‘trusted’ receiver. The EU global satellite
navigationsystemGalileoofferssuchaserviceinits
portfolio. However, the implementation and
operation of the
encrypted services will call for
differentuserequipment.
5 GNSSSPOOFINGDETECTIONAND
MITIGATIONUSINGDISTRIBUTEDGNSSSDR.
Resulting from understanding of GNSS position
estimation process in modern GNSS receivers
(Sections 2 and 3) and the nature of the GNSS
spoofing problem (Section 4), the GNSS spoofing
detection and mitigation (GNSS
SDM) method that
utilisesdistributedGNSSSDR(introducedinSection
3)isproposedinthisSection,anddepictedinFigure
3.
The introduction of the GNSS SDM counter‐
measure relies on the presumptions, as follows: (i)
implementation of distributed GNSS DR concept,
withtherequiredcomputationalandcommunication
capacity fully
operational (Section 3), (ii) seamless
and continuous access to GNSS assistance data
(including streams ofbroadcast navigation message)
from sources such as EGNOS data streaming, or A‐
GNSS service provided with mobile
telecommunication networks(Petrovski and Tsujii,
2012).
The GNSS SDM counter‐measure method is
implemented within the architecture depicted in
Figure
3 using a spoofing detection algorithm
(Algorithm 1, below) based on comparison between
thebroadcast(‘trusted’)andreceived(byuserGNSS
receiver) navigation message. The GNSS SDM
algorithmisdevelopedinorderto identify potential
mis‐matchesinthetwosetsofdata,andtoattemptto
classifythecause
ofsuchamis‐match(lossofdatain
communication, or a pattern that suggest possible
intentionalmanipulationwithdata).
654
Figure3.GNSSanti‐spoofingoperationalmethod,basedon
distributedGNSSSDRreceiverandauxiliaryassistance
Algorithm1:Spoofingdetectionbycomparisonofbroadcast
andreceivednavigationmessages
_______________________________________________
Data:Twoequallydimensioneddataframesb[n,m]andr[n,
m]containingbinarycontentofbroadcast,andreceived
GPSnavigationmessage,respectively.
Result:Dataframeflags[n,m]offlags indicatingequalityof
relatedbitsofbinarycontent
1 readtwodataframesbandr;
2 createemptyresult
dataframeflags[n,m];
3 fori:=1tondo
4 forj:=1tomdo
5 if(b[i,j]==r[i,j]){
6flags[i,j]==1}else{
7flags[i,j]==0}
8 end;
9 end;
10end;
_______________________________________________
Proposed concept, including the GNSS SDM
architectureandalgorithm,wasvalidatedunderthis
study through simulation developed in the open
source R framework for statistical computing. (R‐
project, 2018). Different approaches in data frame
comparisonwereusedthatallowsforidentificationof
mis‐matchesandgenerationofflagmatrixindicating
them.
The R‐based algorithms were based on the
earlyworksinstringcomparison(Levenshtein,1966),
(LandauandVishkin,1988),and(Navarro,2001)and
their implementations either in core R, or in
supported R software libraries, such as: compareDF
andarsenal.TheinitialsuccessfulresultsoftheGNSS
SDM method deployment
have been accomplished,
encouraging further research in advanced GNSS
spoofing detection through statistical learning‐based
mis‐match pattern detection and classification. With
simple statistical learning models already providing
unexpected good performance, the GNSS SDM
method is a potential candidate for successful
detectionandclassificationofvariousformsofGNSS
spoofing
attacks.
Data transfer of RF domainsignal snapshots and
original (broadcast) navigation message may be
performed using the advanced cryptographic
methodstoassurecommunicationsecurity.Quantum
computing and cryptography methods may
considerablyrisethesecuritylevel.Atthesametime,
the analytics of communications may reveal hidden
details of the attacks
declined and of their
perpetrators(Wittek,2014).
Once discovered, cases of GNSS spoofingattacks
maybeexaminedfurtherto identifytheregion(s)of
GNSS spoofing operation, and even to identify the
perpetrator using spatial statistical learning and
location intelligence techniques and methods. GNSS
SDM method is deployed at the cloud
level, thus
capableofaggregationandanalyticsofmassivedata
setswithpotentialsourcesofGNSSspoofingdata.At
thesametime,accessto‘trusted’navigationmessage
allowsforcorrectimplementationandGNSSspoofing
elimination in distributed GNSS SDR position
estimation process. Statistical learning‐based
characterisation, classification and location
intelligencemay
matureinapowerfultooltocombat
GNSS spoofing attacks on the grounds of artificial
intelligence.
Proposed GNSS SDM deployment does not
require any modification of the core GNSS systems,
but the provision of original (broadcast) navigation
message content using thealternative means
(internetstreaming,A‐GNSSetc.).Moreover,modern
mobile platforms with embedded GNSS receivers
(smartphones, Internet‐of‐Things GNSS‐equipped
devices,vehicles,aeroplanes,vesselsetc.)mayrender
transition to distributed GNSS SDR fairly easily. In
sum, the establishment of the environment for the
GNSS SDM method implementation may be
performedseamlesslyandefficiently.
6 GNSSSPOOFINGINMARITIMESECTOR
The problem of GNSS spoofing attacks should be
thoroughly considered as serious threat for GNSS‐
based applications in maritime sector. Research
groups have already demonstrated successful GPS
spoofing attacks in air (Kerns et al, 2014), maritime
(Bhatti,andHumphreys,2017),andland(Zengetal,
2017) navigation applications. Maritime sector
relies
increasingly on information and communication
technologies, with numerous operational procedures
underdevelopmentanddatamanagementnotalways
considered a critical element of sustainable
developmentandoperations.Despitetheallegations,
there has not been confirmations of any real GNSS
spoofingattackinpracticeyet.However,thismaybe
takenasan
advantageonly,providing the breathing
time to develop and validate GNSS anti‐spoofing
measures before the actual GNSS spoofing attack
occurs.
Proposed GNSS SDM is applicable in maritime
sector through various means of implementation.
Improved communication capacity even at the open
seaassuresaccesstoGNSSassistancedata.Compared
with
personal mobile platforms, vessels may host
more capable computational facilities to serve as
mobile hubs for GNSS spoofing detection and
classificationinamobileenvironment,providingthe
GNSS spoofing‐related knowledge on the global
basis.
655
7 CONCLUSION
GNSS spoofing has been recognised a particularly
harmful and serious form of threat to GNSS
performance and operation. Numerous research
groups have worked extensively on the problem.
GNSSspoofingoperationhasbeendemonstratedina
number of research projects, justifying a significant
effortincounter‐measuresdevelopment.
Curiously
enough, only the unconfirmed
allegations of GNSS spoofing inoperation have
beenrecordedsofar.However,thethreatmustnotbe
treatedlightly,butjustasanopportunitytocounter‐
measure the cyber‐attack of potential serious effects
on the economy, safety and security, and society in
general.
Here the
foundations of a novel approach in
combating GNSS spoofing threats are presented.
After a brief outline of GNSS position estimation
process and its shortcomings and vulnerabilities, a
thoroughexaminationoftheGNSSspoofingproblem
isgiven.PotentialsofvariousGNSSspoofingcounter‐
measures development were assessed. The GNSS
Spoofing Detection and
Mitigation (GNSS SDM)
method was developed and outlined, within the
establishedframeworkarchitectureandrequirements
for GNSS anti‐spoofing. The proposed method does
notrequire modificationof either the existing GNSS
coresystems,ortheprevailingmajorityoftheexisting
user equipments (smartphones, in particular). The
proposed concept is validated
in simulation‐based
scenario developed within the R framework for
statistical computing, demonstrating GNSS spoofing
detectionand statistical learning‐basedclassification.
GNSS spoofing mitigation was conducted using the
verynatureoftheproposedconcept.
GNSS spoofing is a serious information security
threat that should be mitigated efficiently and
completely.Potentialdamage
andliabilitiesresulting
fromGNSSspoofingareenormous.Itisbelievedthat
the proposed method, based on statistical learning,
and potentially enhanced further with cryptography
methods,mayprovideasuccessfultoolincombating
GNSSspoofing.
REFERENCE
Alvarez, G, and Li, S.(2006). Some Basic Cryptographic
Requirements for Chaos‐Based Cryptosystems. Int J of
BifurcandChaos,16(8),2129‐2151.
Babu,AM,andSingh,KJ.(2013).PerformanceEvaluation
of Chaotic Encryption Technique. Amer J of Appl Sci,
10(1),35‐41.
Bhatti, J, and Humphreys, T E.
(2017). Hostile Control of
Ships via False GPS Signals: Demonstration and
Detection. NAVIGATION: Journal of the Institute of
Navigation,64(1),51‐66.doi:10.1002/navi.183
Capparra, G et al. (2016). Navigation Message
AuthenticationSchemes.InsideGNSS,Sep‐Oct2016,64‐
73.
Filić,M,and Filjar,R. (2018a). Forecasting model
of space
weather‐drivenGNSSpositioningperformance.Lambert
AcademicPublishing.Riga,Latvia.
Filić, M, Filjar, R. (2018b). Smartphone GNSS positioning
performance improvements through utilisation of
Google Location API. Proc of 41 st International
ConventionMIPRO/CTI,507‐510.Opatija,Croatia.doi:
10.23919/MIPRO.2018.8400087
Filić,M, Grubišić,L,and Filjar, R.
(2018). Improvement of
standard GPS position estimation algorithm through
utilization of Weighted Least‐Square approach. Proc of
11th Annual Baška GNSS Conference, 7‐19. Baška, Krk
Island, Croatia. Available at:
https://www.pfri.uniri.hr/web/hr/dokumenti/zbornici‐
gnss/2018‐GNSS‐11.pdf
Filić,M.(2017).Analysisofthepositionestimationprocedure
based on the given GNSS pseudoranges in the
Software‐Defined Radio satellite navigation receiver
(Mscthesis,inCroatian).Universityof Zagreb,Faculty
of Science, Dept for Mathematics. Zagreb, Croatia.
Availableat:https://bit.ly/2DlX9O6
Gallier,J,andQuaintance,J.(2018).FundamentalsofLinear
Algebra and Optimization. Dept of Computer and
Information Science, University of Pennsylvania.
Philadelphia, PA. Avaialble at:
https://www.seas.upenn.edu/~cis515/linalg.pdf
Gustafsson, F. (2010). Statistical Sensor Fusion.
Studentlitteratur.Linkoepping,Sweden.
Humphreys,Tetal.(2008).AssessingtheSpoofingThreat:
DevelopmentofaPortableGPSCivilianSpoofer.Procof
IONGNSS2008,3584‐3590.Savannah,GA.
Jafarnija‐Jahromi, A et al. (2012). GPS Vulnerability to
Spoofing Threats and a Review of Antispoofing
Techniques.IntJofNavandObs,2012,ArticleID127072,
16pages.doi:10.1155/2012/127072
Kerns,AJ,Shepard,DP,Bhatti,JA, andHumphreys,TE.
(2014a). Unmanned Aircraft Capture and Control via
GPS Spoofing.Jof Field Robotics,July/Augist2014,617‐
636.doi:10.1002/rob.21513
Kerns,AJ,
Wesson,KD,andHumphreys,T E.(2014b).A
Blueprint for Civil GPS Navigation Message
Authentication.ProcofIEEE/IONPLANS2014(8pages).
Monterey,CA.
Kuhn, M G. (2010). Signal Authentication in Trusted
SatelliteNavigationReceivers.Chapterin:Sadegi,A‐R,
and Naccache, D (eds). (2010). Towards Hardware‐
Intrinsic Security: Foundations
and Practice. Springer
Verlag.NewYork,NY.
Landau,GM,andVishkin,U.(1988).FastStringMatching
withkDifferences.JCompandSystSci,37,63‐78.
Levenshtein, V I. (1966). Binary Codes Capable of
Correcting Deletions, Insertions, and Reversals
(translated from Russian version). Cybernetics and
ControlTheory,10(8),
707‐710.
Mitola, J. (1995). The Software Radio Architecture. IEEE
CommMag,May1995,26‐38.
Navarro,G.(2001).AGuidedTourToApproximateString
Matching.ACMCompSurveys,33(1),31‐88.
Nighswander,Tetal.(2012).GPSSoftwareAttacks.Procof
CCS2012Conference.Raleigh,NC.
O’Driscoll, C.
(2018). What is navigation message
authentication? Inside GNSS, Jan/Feb 2018, 26‐31.
Availableat:https://bit.ly/2SEM6UC
Pecora,Metal.(1997).Fundamentalsofsynchronisationin
chaotic systems, concepts, and applications. Chaos, 7,
520‐543.
Petrovski, I G, and Tsujii, T. (2012). Digital Satellite
Navigation and Geophysics: A Practical Guide with
GNSS
Signal Simulator and Receiver Laboratory.
CambridgeUniversityPress.Cambridge,UK.
Psiaki, M L, and Humphreys, T E. (2016). GNSS Spoofing
andDetection.ProcofIEEE,104(6),1258‐1270.
Schenewerk, M et al. (2016). Quality Controlling RINEX
Navigation Message Files. Proc IGS Workshop: GNSS
Futures. Sydney, NSW. Available at:
https://bit.ly/2RwUvIu
Stewart R W et al. (2015). Software Defined Radio using
MatLab & Simulink and the RTL‐SDR. Strathclyde
AcademiaMedia.Glasgow,UK.
656
Sun, K. (2015). Chaotic Secure Communication: Principles
and Technology. Walter De Gruyter GmbH. Berlin,
Germany.
Tippenhauer, N O, Poepper, C, Rasmussen, K B, and
Čapkun, S. (2011). On the Requirements for Successful
GPSSpoofingAttacks.Procofthe18
th
ACMconferenceon
Computerandcommunicationssecurity,75‐86.Chicago,IL.
UKGovernmentOfficeforScience. (2018).Satellite‐Derived
Time and Position: A Study of Critical Dependencies.
HM Government of the UK and NI. Available at:
https://bit.ly/2E2STnd
Vaudenay, S. (2005). A Classical Introduction to
Cryptography: Applications for Communications
Security.Springer
Verlag.NewYork,NY.
Wesson, K, Rothlisberger, M, and Humphreys, T. (2012).
Practical Cryptographic Civil GPS Signal
Authentication. NAVIGATION: Journal of the Institute of
Navigation,59(3),177‐193.
Wittek, P. (2014). Quantum Machine Learning: What
Quantum Computing Means to Data Mining.
AcademicPress/Elsevier.London,UK.
Zeng, K et al.
(2017). A Practical GPS Location Spoofing
Attack in Road Navigation Scenario. Proc of HotMobile
2017Conference(6pages).Sonoma,CA
