Measurement Science and

Technology

PAPER You may also like

- Face Anti-Spoofing Method with Blinking
GNSS spoofing detection technology based on Eye and HSV Texture Analysis
Raden Budiarto Hadiprakoso
Doppler frequency shift difference correlation - Carrier phase double-difference GNSS
spoofing detection technology based on
generalized likelihood ratio
To cite this article: Junzhi Li et al 2022 Meas. Sci. Technol. 33 095109 Jidong Yin, Le He and Xuyang Lai

- Performance analysis of generated

spoofing under integrated navigation
conditions
Menghao He, Changgeng Li, Chao Zhou
View the article online for updates and enhancements. et al.

This content was downloaded from IP address 106.39.42.200 on 17/09/2023 at 05:22

 Measurement Science and Technology

Meas. Sci. Technol. 33 (2022) 095109 (11pp) https://doi.org/10.1088/1361-6501/ac672a

GNSS spoofing detection technology

based on Doppler frequency shift
difference correlation
Junzhi Li1, Xiangwei Zhu1,∗, Mingjun Ouyang1, Dan Shen2,
Zhengkun Chen1 and Zhiqiang Dai1
1
School of Electronics and Communication Engineering, Sun Yat-Sen University, Shenzhen,
People’s Republic of China
2
School of Systems Science and Engineering, Sun Yat-Sen University, Guangzhou, People’s Republic
of China

E-mail: [email protected]

Received 3 January 2022, revised 27 March 2022

Accepted for publication 13 April 2022
Published 27 June 2022

Abstract
Global navigation satellite system (GNSS) spoofing interference poses a great threat to the
security of end users. Based on the same propagation path characteristics of the spoofing signals
transmitted by a single antenna spoofing device, this paper proposes a method against single
antenna spoofing jamming utilizing the Fréchet distance of Doppler frequency difference. When
the receiver moves, the Doppler data in a window are fitted via the leastsquares method, and the
Fréchet distance between the two satellite signals is calculated to obtain the similarity evaluation
value between them. The detection of spoofing signal is carried out based on the evaluation
value. We use real data to verify the detection performance of the method, and the experimental
results show that the method can effectively distinguish the spoofing signal from the authentic
signal. In addition, the method has low computational complexity and requires less additional
information, which makes it possible to apply it in the anti-spoofing module of a GNSS receiver.
Keywords: GNSS, spoofing jamming, Doppler frequency difference, Fréchet distance,
spoofing detection

(Some figures may appear in colour only in the online journal)

1. Introduction which had gradually focused the public’s vision on the safe
application of GNSS [3].
Nowadays, global navigation satellite systems (GNSS) play an The GNSS vulnerability can be analyzed from the devel-
increasingly important role in the military and civil fields. The opment process of a satellite navigation system. There are
wide application of GNSS provides stable and reliable space- two reasons for the GNSS vulnerability. On the one hand,
time benchmark information for various fields, bringing signi- the GNSS orbit design adopts the Walker constellation based
ficant scientific and technological changes to many important on medium Earth Orbit satellites, and even selects high orbit
fields [1]. In the past, the end users were more interested in the satellites such as the inclined geosynchronous satellite orbit
availability and accuracy of GNSS navigation, positioning and and geosynchronous Earth Orbit satellites. Due to the lim-
timing, and ignored the aspect of security [2]. In recent years, ited resources and energy of satellite platforms, the GNSS
spoofing jamming attacks have occurred continuously. This cannot transmit highpower signals in the complete airspace.
exposes the vulnerability of the GNSS signals to interference, Moreover, it is far away from the ground users, and the trans-
mitted signals suffer huge losses in the transmission path [4].
On the other hand, direct sequence spread spectrum tech-
∗
Author to whom any correspondence should be addressed. nology based on code division multiple access is usually used

1361-6501/22/095109+11$33.00 Printed in the UK 1 © 2022 IOP Publishing Ltd

Meas. Sci. Technol. 33 (2022) 095109 J Li et al

in the GNSS signal system design. It exploits the pseudoran- by observing the estimated receiver clock state. The detection
dom code sequence to significantly increase the bandwidth of performance is well analyzed in the paper, but it is lack of real
transmitting signal compared to that of the information signal. spoofing scence verification. Akos [14] studied the automatic
To facilitate signal acquisition, tracking, message analysis and gain control monitoring method. This method can not suppress
other operations by the user’s receiver, it is necessary to pub- the influence of spoofing signal, and has little effect on some
licly release the signal structure, debugging mode, message spoofing attacks. Jafarnia et al [15] proposed the anti-spoofing
arrangement and other information of the GNSS signal in the method based on absolute power and noise level monitoring.
form of interface control document. This greatly promotes the However, this method is vulnerable to the influence of mul-
popularization of GNSS, but also creates the hidden danger tipath propagation, antenna type and attitude, resulting in false
that the receiver may receive false signals. Even the military alarm, which makes its detection accuracy not high. Psiaki
signals that need authorization can experience spoofing jam- et al [16] proposed a method of receiver autonomous integrity
ming by forwarding spoofing jamming [5]. monitoring based on pseudorange residuals. Gross et al and
Nowadays all sectors of the society are concerned with the Hu et al [17, 18] studied the spoofing detection method based
security problem of GNSS signal that is vulnerable to interfer- on signal quality monitoring. However, multipath interference
ence. The sectors especially concerned are the unmanned sys- can easily affect its accuracy.
tems, power, communications and other industries that require Next, we review anti-spoofing technology based on signal
high system security and reliability. The demand for trusted space characteristics. Cho et al [19] proposed a method for
navigation capability is increasing day by day with the rapid spoofing detection based on signal arrival time. This method
development of unmanned systems and other fields. Ensuring is generally more effective for forwarding spoofing jamming,
the safe application of GNSS signal has become a hot topic in which is not suitable for detecting generative spoofing jam-
the academic and industrial circles [6]. ming signal. It will even eliminate the authentic signal and
Spoofing jamming is a more complex and covert attack retain the spoofing signal. In addition, it is easy to misjudge
method compared with traditional jamming such as suppres- the multipath signal. Psiaki et al and Jafarnia et al [20, 21]
sion jamming. It can cheat the receiver to the wrong time studied anti-spoofing technology based on multi-antenna DOA
and position by transmitting false GNSS signals without being detection. However, this method has high requirements for the
detected by the terminal equipment. Therefore, spoofing has a receiver and needs to be equipped with array antenna, which
higher risk factor than interference [7]. At present, the research is too expensive for civil receivers. Ni et al [22] proposed a
on anti-spoofing countermeasures against spoofing jamming spoofing detection method based on carrier phase double dif-
attack is continuously emerging. ference observation for the detection and suppression of air-
In terms of signal encryption mechanism, the authors in craft spoofing interference. The detection of spoofing signal
[8] studied the method of using spread spectrum code encryp- is realized by comparing the double difference observation
tion to resist spoofing jamming. Humphreys [9] proposed a results of two different signals with the measurement error
navigation message authentication strategy to resist spoofing output of the receiver. However, this method needs to further
attacks, and designed a security code estimation and replay optimize and solve the problem of cycle ambiguity in order
spoofing detection test for encrypted GNSS signals. However, to better ensure the detection effect. Humphreys et al [23]
the statistical model designed in this experiment is effect- proposed a residual signal detection method for anti-spoofing.
ive only when there are only real signals or spoofing sig- However, when the power of the spoofing signal is high, the
nals. When testing in the presence of spoofing signal and authentic signal may be suppressed under the noise, result-
authentic signal at the same time, although it has a certain ing in the receiver capturing the spoofing signal incorrectly.
detection effect, the detection sensitivity is relatively low, Therefore, the application scenario of this method is limited,
and the performance needs to be further optimized. Wesson but it can be detected jointly with the power anomaly detec-
et al [10] proposed a practical civil GPS signal authentica- tion method. Broumandan et al [24] studied a GNSS spoof-
tion technology that combined encryption authentication of ing detection method based on signal spatial correlation. This
GPS navigation information with signal timing authentica- method is suitable for detecting the interference signal sent by
tion based on the statistical hypothesis test. However, the nav- the single antenna spoofing equipment. However, this paper is
igation information authentication (NMA) strategy of nav- only theoretical analysis and does not use real data to test its
igation message is only for civil L2/L5 frequency points. performance.
Kerns et al [11] also proposed a scheme of NMA implemen- Other references studied the aspect of spoofing detection
ted on modern civil GPS signals. This study only provided technology based on external devices. Lee et al [25] proposed
the possibility of realizing civil NMA strategy and detailed an accelerometer assisted anti-spoofing method. This paper is
strategy blueprint, which has not been verified in the actual mainly theoretical analysis, and has not been tested and veri-
scenario. However, it is difficult to implement this techno- fied in the real scene. James and Broumendan [26] proposed
logy in a short time due to the cost and complexity of signal a low-cost spoofing detection strategy for an inertial meas-
encryption. urement unit (IMU). However, IMU devices with different
Based on the data characteristics anomaly caused by spoof- accuracy levels have a certain impact on the detection perform-
ing attacks, the anti-spoofing research also has more inform- ance, and its performance needs to be further verified. Liu et al
ation. In [12, 13], a receiver autonomous signal authentica- [27] studied a strategy of combining inertial navigation system
tion method was proposed, which detected spoofing signals and GNSS system, and proposed a GNSS spoofing detection

2
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

method based on the Kalman filter. This method does not make
an in-depth analysis of its detection performance and lacks the
verification of real data.
Research has also been carried out on spoofing detection
technology based on deep learning. Semanjski et al and Koji
[28, 29] studied, analyzed and verified the GNSS signal spoof-
ing detection method based on supervised machine learning.
Li et al [30] used the CNNs to study the problem of spoofing
signal detection in the acquisition phase. Although the spoof-
ing detection method using deep learning has a certain detec-
tion effect, it is mostly based on a certain spoofing scene mode.
The model trained through the data set is relatively solidified.
Once the data in different scenes changes, it may lead to failure
and lack of self adaptability.
In this paper, based on the characteristics of spoofing sig-
nals transmitted by a single antenna spoofing jamming source,
a spoofing detection method utilizing the Fréchet distance Figure 1. Spatial sampling diagram of the antenna trajectory of
of Doppler frequency shift difference (DFSD) of a mobile receiver.
receiver is proposed. The DFSD data are obtained after fit-
ting the Doppler data. By comparing the DFSD data between
two signals using the Fréchet distance method, the spoofing However, the propagation paths between the spoofing sig-
signal and the authentic signal can be detected and recog- nals remain the same, and the Doppler frequency shift has
nized. Simulation and real data are used to verify the effect- a similar change rule with a high correlation, as shown in
iveness of the method. This method has low complexity and figure 1.
low cost, because the technology does not need to modify the In this paper, we only deal with the Doppler frequency
receiver hardware, and does not need other external informa- difference data based on a PLL. This choice avoids complex
tion except the measured Doppler frequency. It can not only model information such as Doppler and receiver clock drift
detect spoofing signal, but also distinguish spoofing signal caused by the satellite motion, and also lowers the complexity
from real signal. of the algorithm. Without considering the satellite ephemeris,
receiver position and clock drift model, the similarity between
two signals is estimated by calculating the Fréchet distance
2. Problem description between them. Other conditions ignored in this process will
not affect the detection and recognition performance of the
Spoofing jamming is a kind of jamming with concealment and algorithm.
attack. Based on the different methods of generating spoofing
signals, the spoofing jamming can be divided into forward- 3. Description of the proposed method
ing spoofing jamming and generating spoofing jamming [31].
The existing spoofing jamming methods usually transmit the When the receiver successfully tracks the satellite signal, the
spoofing signal to the target area through a single antenna to measured Doppler frequency can be expressed as follows,
carry out a spoofing attack. Due to a single antenna being used based on the signal processing mechanism of the receiver:
to transmit signals, the technical and equipment costs are rel-
atively low, and the method is easy to implement. However, fd (t) = fds (t) + fdr (t) + fe (t) + δfu (t) − δfs (t) + Tr(t)
the spoofing jamming mode of multi-antenna high coopera-
tion has high technical requirements, which not only requires − I(t) + ε(t) (1)
cooperation between signals, but also requires a high degree of
clock synchronization. In addition, the target receiver is often where fd (t) is the measured Doppler frequency, fds (t) and fdr (t)
mobile, which makes it extremely difficult and costly to realize are the Doppler caused by the satellite and receiver motion,
this type of spoofing jamming [32]. respectively. fe (t) represents the Doppler caused by the earth’s
The signal from a single antenna has similar spatial char- rotation. δfu (t) and δfs (t) are the receiver and satellite clock
acteristics, which is also the main idea used in this paper. drifts. The tropospheric delay rate and the ionospheric delay
As different satellite signals transmitted by the same antenna rate are denoted by Tr(t) and I(t), respectively, and ε(t) is the
have the same propagation paths, they have almost the measurement Gaussian white noise.
same channel gains and power spectral densities for differ- We can obtain the Doppler frequency of each signal from
ent spacetime points. As the receiver moves, the propaga- the receiver tracking loop. Subsequently, we can get the Dop-
tion path between the real satellite signals constantly changes, pler numerical sequence [ fdi (1), fdi (2), . . . , fdi (n), . . . , fdi (N)],
resulting in the change of doppler frequency shift, which where fdi (n) represents the measured Doppler frequency of the
decreases the correlation between different satellite signals. i-th satellite signal at time n × ∆T, and ∆T is the sampling

3
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

4. Methodology

Next, we introduce the method for evaluating the similarity of

two groups of data. There are many direct and effective meth-
ods for similarity evaluation, such as distance evaluation based
on various distance measures, similarity analysis using correl-
ation coefficient, and so on. Out of these methods, the Euc-
lidean and Hausdorff distances are widely used. In this paper,
we use the Fréchet distance method, which can describe the
spatial path similarity. This method considers the path spatial
distance, so that it can efficiently evaluate the curve similarity
with a certain spatial time series.
The mathematical definition of Fréchet distance [33, 34] is
as follows:
Let two tuples (S, d) represent a metric space, where d is a
metric function on S. Without specifying the metric function,
we call the metric space S for short. Let A and B be two con-
Figure 2. The least square model fitting Doppler frequency data. tinuous curves on S, i.e. A: [0,1] → S and B: [0,1] → S. Let
α and β be two multiparametric functions of the unit inter-
val, i.e. α:[0,1]→S and β:[0,1]→S. Then, the Fréchet distance
interval. In a short period of time, we can consider the Dop- between curves A and curve B are defined as
pler numerical sequence to change linearly. Consequently, we
can use the least-squares linear regression model to fit it as F(A, B) = inf max {d(A(α(t)), B(β(t)))} (4)
α,β t∈[0,t]
shown below:
where Inf represents the lower bound of the value.
fgi (n) = â × n + b̂ (2) The above expression represents the continuous Fréchet
distance formula. However, as our data are discrete, we should
where fgi (n) is the Doppler frequency estimated by the model, apply the discrete Fréchet distance formula to the data. We use
Euclidean distance to calculate the distance between discrete
â and b̂ are the estimated model parameters, and n is the ordinal
points.
of the sequence. Figure 2 shows the fitting curve.
The discrete Fréchet distance can be transformed as fol-
We can get the estimated sequence [ fgi (1), fgi (2), . . . ,
lows: In the unit interval [0,1], randomly extract (n + 2) mutu-
fg (n), . . . , fgi (N)] based on the above fitting. Now, after
i
ally different numbers to form a monotone sequence (tk )nk= +1
0,
subtracting the sequence value of the least squares such that t0 = 0, tn+1 = 1, and tk < tk+1 . Subsequently, the dis-
linear estimation from the real Doppler sequence, cretization formula of Fα,β (A, B) can be expressed as
we can obtain a Doppler shift difference sequence
[∆f i (1), ∆f i (2), . . . , ∆f i (n), . . . , ∆f i (N)]. ⌢
F nα,β (A, B) = max {d(A(α(t)), B(α(t)))} . (5)
Therefore, ∆f i (t) can be expressed as +1
t∈(tk )nk= 0

∆f i (n) = fdi (n) − fgi (n). (3) Then the discretization formula of F(A, B) can be
expressed as
As the spoofing signals are broadcast from the same ( )
antenna, the position vector between the receiver and the F(A, B) = inf max {d(A(α(t)), B(β(t)))} . (6)
antenna is the same and, therefore, the propagation path of the α,β +1
t∈{tk }nk= 0

spoofing signal to the receiver is the same. This will cause

the Doppler shift difference between any two spoofing sig- In fact, the discrete Fréchet distance approximates the con-
nals to be almost the same, i.e. ∆f i (t) and ∆f j (t) will be tinuous Fréchet distance. When there are enough discrete
almost equal. The experimental results in section 5 also prove points on the curve, the discrete and continuous Fréchet dis-
this result. tances are approximately equal. Note that the smaller the
In a real scene without spoofing, if the receiver moves ran- Fréchet distance, the higher the similarity between the two
domly, the DFSD between any two authentic signals is not the groups of data, and vice versa.
same because the position vectors between the receiver and the Next, we calculate the similarity between the DFSD data
satellites are not the same. Hence, the similarity between two of any two signals obtained in the previous section. First, we
authentic signals is low. Based on this feature, we can detect can use the sequence window method to divide the processed
and recognize the authentic and spoofing signals by compar- Doppler frequency shift sequence into many subsequences
ing the similarity of DFSD between them. [∆f i (1), ∆f i (2), . . . , ∆f i (k), . . . , ∆f i (K)] of equal length.

4
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

The sub-sequences can be expressed as

[∆f i (1 + (k − 1) × M), ∆f i (2 + (k − 1) × M), . . . ,

∆f i (n + (k − 1) × M), . . . , ∆f i (k × M)]

where M is the length of the subsequence. Thus, the Fréchet

distance between the i-th signal and the j-th signal can be
obtained as F(∆f i (k), ∆f j (k)).
The sample standard deviation of a subsequence can be cal-
culated as
v
u
u 1 X M
σk (M) = t
2
i
(∆f i (n + (k − 1) × M) − ∆f̄ki ) . (7)
M−1
n=1 Figure 3. Schematic diagram of angle setting between move
receiver and satellite.
There will be two hypotheses for classification, depend-
ing on the presence of spoofing signals, and the presence of
only authentic signals. The zero assumption is that there is In the simulations, the receiver moves randomly at a cer-
no spoofing jamming, which corresponds to the case where tain speed and in different directions. In addition, the elev-
the Fréchet distance of DFDS is greater than the threshold. ation angle of the satellite is randomly generated according
Another assumption is that there is spoofing jamming, and the to the uniform distribution U(0, π/ 2). The direction angle is
Fréchet distance corresponding to the DFDS is less than the also randomly generated according to the uniform distribu-
threshold. These two hypotheses can be described as follows: tion U(−π/ 3, π/ 3), which is based on the maneuverability
 of most vehicles. The above settings change once per second.
H0 : F(∆f i , ∆f j ) < Th Spoofing signals The SNR and the speed of the receiver are fixed at 10 dB and
. (8)
H1 : F(∆f i , ∆f j ) ⩾ Th Authentic signals 10 m s−1 , respectively. In the experiments, we generate six
A large number of simulation and real data experiments satellite signals, which are named as PRN 6, PRN 10, PRN
show that there is a certain relationship between the threshold 12, PRN 17, PRN 20 and PRN 23. Then, there will be a total
and the standard deviation of the data. In this paper, we set of 15 combinations of satellite signals.
a dynamic threshold according to the state of the data in the In the real scene experiment, we generate the spoofing
window, which is expressed as signal through the spoofing signal analog source. When the
spoofing signal is synchronized with the authentic signal, the
Th = ζ(σMi + σMj ) (9) spoofing signal is broadcast, and finally the data is collected
through the leader receiver, and the Doppler frequency in the
where σMi and σMj are the standard deviation of the data in the original observation information is output. In the process of
window of the i-th signal and the j-th signal respectively. ζ is signal acquisition, considering that the user receiver is unlikely
the threshold weight, which can be adjusted according to the to move at a fixed speed all the time, we complete the acquisi-
window width. tion of real scene data by continuously and randomly moving
Next, we will verify the detection performance of the pro- the handheld receiver at different speeds. This is also relat-
posed method through simulation experiments and real data. ively more in line with the movement of the end user of the
receiver. Figure 4 shows the scene where we collect spoofing
5. Experiments and data analysis signals and deception jamming source equipment.
In addition, considering the relationship between the
5.1. Simulation settings threshold and the standard deviation, the sample standard devi-
ation should be made as close to the real data standard devi-
We need to adopt a simple and effective simulation analysis ation as possible. This requires the length of the window
method combined with practice. In this paper, we only con- sequence to be large enough. However, when the length is
sider the influence of receiver motion. This is because the Dop- too large, it will lead to a high computational complexity and
pler caused by satellite motion in a short time can be con- increase the detection time. The length can be adjusted accord-
sidered as linear, and the influence of satellite motion can be ing to the sampling rate of the receiver. In order to ensure the
approximately eliminated after the least squares fitting. validity of the dynamic threshold, the data length can be within
Figure 3 is a schematic diagram of the angle between the [50–200]. In the experiments, the length of window sequence
moving directions of the receiver and the satellite. We set a is set as M = 50, the ζ value is set to 4.
reference direction. The angle between the motion direction of
the receiver and the reference direction is denoted by α. The
5.2. Simulation experiments and analysis
azimuth of the satellite PRN-i and the azimuth of the satellite
PRN-j are represented by αi and αj , respectively. The eleva- Figure 5 shows the Doppler shift difference data of the sim-
tions of the satellites PRN-i and PRN-j are denoted by θi and ulated satellite and spoofing signals at the receiver speed of
θj , respectively. 10 m s−1 . It can be observed from figure 5(a) that the Doppler

5
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

Figure 4. Real experimental scenes and spoofing source devices.

Figure 5. Doppler frequency difference data of simulated satellite signals and spoofing signals. Simulate authentic signals, (b) simulate
spoofing signals.

Figure 6. Fréchet distance corresponding to Doppler shift difference in figure 5. (a) Simulate authentic signals, (b) simulate spoofing
signals.

shift difference data of satellite signals are disordered and the Figure 6 shows the Fréchet distance corresponding to the
coincidence degree is low. On the other hand, the Doppler shift Doppler frequency difference in figure 5. It is obvious that the
difference data of deceptive signals in figure 5(b) are highly Fréchet distance between any two simulated authentic satel-
coincident, showing obvious similarity between the signals. lite signals is larger than that between the spoofing signals. By
This is consistent with the analysis carried out in section 2. comparing the results of figures 5 and 6, it is easy to distinguish

6
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

Figure 7. Fréchet distance of Doppler frequency difference between

Figure 9. Fréchet distance of Doppler frequency difference of
signals under different sequence lengths (a) Authentic signals, simulate authentic signals.
(b) spoofing signals.

Figure 8. Fréchet distance of Doppler frequency difference Figure 10. Fréchet distance of Doppler frequency difference of
between signals under different SNR. simulated spoofing signals.

the characteristic differences between the satellite and spoof- simulated satellite and spoofing signals decreases as the SNR
ing signals. Thus, the detection and recognition of spoofing increases. However, the Fréchet distance of the simulated
and authentic signals can be realized by setting a reasonable satellite signals is always higher than that of the spoofing
threshold. signals.
Figure 7 shows the relationship between the sequence Figure 9 shows the Fréchet distance of Doppler frequency
lengths of different windows and the Fréchet distance of difference in individual epochs of the simulated satellite sig-
the Doppler frequency difference. As the sequence length nals under SNR = 10 dB. The figure shows that in each epoch,
increases, the distance difference gradually increases. How- the Fréchet distance of the two satellite signals exceeds the
ever, the Fréchet distance of the Doppler frequency difference detection threshold. This indicates that all these signals are
of the authentic signals are significantly higher than that of the recognized as satellite signals.
spoofing signals, which is consistent with the results shown in Figure 10 shows the Fréchet distance of the Doppler fre-
figure 6. quency difference of the simulated spoofing signal when the
Figure 8 shows the Fréchet distance between the simu- SNR is 10 dB. The figure shows that in each epoch, the Fréchet
lated authentic and spoofing signals under different SNR val- distance between any two spoofing signals is lower than the
ues. The figure shows that the Fréchet distance between the threshold value. This behavior indicates that these signals are

7
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

Figure 11. Detection probability under different SNR.

Table 1. Detection probability under different conditions.

Detection probability
−1 −1 −1
SNR At 1 m s At 5 m s At 10 m s At 15 m s−1 At 20 m s−1 At 25 m s−1 At 30 m s−1

5 dB 0.1849 0.2061 0.3289 0.4324 0.5161 0.5776 0.6199

8 dB 0.6647 0.9844 0.9935 0.9972 0.9992 0.9998 1.0
10 dB 0.6754 0.9998 1.0 1.0 1.0 1.0 1.0

spoofing signals. The results of figures 9 and 10 again verify data through a serial port. The real satellite signal data are col-
the difference between spoofing and satellite signals. At the lected in the airspace without any spoofing signal. In the spoof-
same time, the effectiveness of the algorithm is confirmed. ing jamming scene, we use the spoofing jamming source to
Figure 11 shows the detection probability under different broadcast the spoofing signals, and collect them through the
SNR when the moving speed is 5 m s−1 . From the figure that receiver equipment. In order to ensure the successful spoof-
when the SNR is greater than 2 dB, the detection probabil- ing effect, we set the mode that the spoofing signal power is
ity gradually increases with the increase of the SNR, and the about 10 dB higher than the authentic satellite signal power to
detection performance can reach 100% at 10 dB. This indicates implement the spoofing attack.
that the change of SNR has a great influence on the detection Figure 12 shows the Doppler frequency difference of the
probability under the same conditions. authentic and spoofing signals obtained after processing the
Table 1 shows the detection probability under different collected data. Figure 12(a) shows that the Doppler frequency
SNR values and at different moving speeds. It can be noted difference of each authentic signal is obvious and has a low
from the table that the moving speed and the SNR influence similarity. On the other hand, the Doppler frequency differ-
the detection probability. However, if the moving speed of the ence data between spoofing signals shown in figure 12(b) are
receiver is too small, such as 1 m s−1 , the Doppler frequency highly coincident and have high similarity. Figure 13 shows
caused by the movement of the receiver is also very small, the Fréchet distance corresponding to Doppler frequency dif-
which results in a low detection probability. Therefore, in the ference of data in figure 11.
follow-up study, an appropriate speed should be selected. Figure 14(a) shows the Fréchet distance of Doppler fre-
quency difference calculated from 120 s of the authentic satel-
lite signal data. It is clear from the figure that the Fréchet
5.3. Real data experiments and analysis
distance between the two signals exceeds the set dynamic
To objectively show the performance of the proposed threshold. It shows that these real signals can be detected and
algorithm, we use the Taidou D303-3 receiver to acquire the recognized in each epoch within 120 s. Figure 14(b) shows
authentic satellite and spoofing signals, and verify them using the Fréchet distance of the Doppler frequency calculated from
real data. The data are collected in a spoofing scene and real 150 s of the spoofing signal data. The figure shows that in
scene. Our data acquisition platform moves randomly during the 150 s epoch, the Fréchet distance between any two sig-
the signal acquisition process, and outputs Doppler frequency nals is below the set dynamic threshold. These results show

8
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

Figure 12. Doppler frequency difference data of authentic satellite signals and spoofing signals.

Figure 13. Fréchet distance corresponding to Doppler shift difference in figure 12. (a) Authentic signals, (b) spoofing signals.

Figure 14. Fréchet distance of Doppler frequency shift difference in each epoch. (a) Authentic signals, (b) spoofing signals.

9
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

Figure 15. Hausdorff distance of Doppler frequency shift difference in each epoch.

that the spoofing signals are detected and accurately identi- Data availability statement
fied in each epoch. By comparing real data in figures 14(a)
and (b) when the receiving terminal moves randomly, where The data generated and/or analysed during the current
the spoofing signal is transmitted by the single antenna spoof- study are not publicly available for legal/ethical reasons but
ing source, it can be observed that the spoofing signal can be are available from the corresponding author on reasonable
accurately identified by detecting the Fréchet distance of Dop- request.
pler frequency difference.
To make the performance more convincing, this paper com- Acknowledgments
pares the performance with the detection method based on
Hausdorff distance. In the experiment, the same authentic sig- The authors thank the associate editor and the reviewers
nal data and spoofing signal data as in figure 15 are used fo rtheir useful feedback that improved this paper. Thanks
for verification, so as to objectively evaluate the performance Changsha Technology Research Institute of Beidou Industry
of different methods. The specific experimental results are Safety for its support of deception simulation source equip-
shown in figure 15. By comparing figures 15(a) and (b) with ment and experimental site.
figures 14(a) and (b), it can be seen that the detection effect of
Hausdorff distance method is not ideal. In some data epochs, Funding
spoofing signals cannot be effectively detected and recog-
nized, and even the authentic signals are mistakenly regarded This research was funded by the Key Special Projects of the
as spoofing signals. The comparison shows that the detection National Key Research and Development Plan of China (Grant
performance of Fréchet distance method is better than that of No. 2018YFB0505200), in part by the National Natural Sci-
Hausdorff distance method. ence Foundation of China (Grant No. 61973328), in part by the
Project supported by the Major Research Plan of the National
6. Conclusions Natural Science Foundation of China (Grant No. 91938301),
and in part by the Key Basic Research Projects of Shenzhen
The GNSS is an essential component in various applications, Science and Technology Commission (Grant No. 2020N259).
therefore, it is important to ensure its safe and accurate usage.
ORCID iDs
This paper proposed a GNSS spoofing detection method based
on a single antenna mobile antenna. The results showed that Junzhi Li  https://orcid.org/0000-0001-7811-4575
the spoofing signal generated by a single antenna spoofing Xiangwei Zhu  https://orcid.org/0000-0002-4415-0698
jamming source could be effectively detected by studying the Mingjun Ouyang  https://orcid.org/0000-0002-4289-3252
spatial correlation of the signal. This method transformed the
spoofing detection problem into a similarity detection prob-
lem, and its detection performance was not affected by the References
spatial propagation multipath fading of the GNSS signal. In
[1] Schmidt E, Ruble Z, Akopian D and Pack D 2019
addition, this method had low complexity and cost, and did Software-defined radio GNSS instrumentation for spoofing
not require any additional estimation techniques, thus making mitigation: a review and a case study IEEE Trans. Instrum.
it convenient to apply it to a hardware system. Meas. 68 2768–84

10
Meas. Sci. Technol. 33 (2022) 095109 J Li et al

[2] Khalajmehrabadi A, Gatsis N, Akopian D and Taha A 2018 [18] Hu Y, Bian S and Cao K 2018 GNSS spoofing detection based
Realtime rejection and mitigation of time synchronization on new signal quality assessment model GPS Solut. 22 1–13
attacks on the global positioning system IEEE Trans. Ind. [19] Cho S, Mi S and Lim S 2008 Design of a TOA-based
Electron. 65 6425–35 anti-spoofing method for GPS civil signal Int. Symp. on
[3] Margaria D, Motella B, Anghileri M, Floch J and Paonni M GPS/GNSS pp 438–49
2017 Signal structure-based authentication for civil GNSSs: [20] Psiaki M, Powell S and O’Hanlon B 2013 GNSS spoofing
recent solutions and perspectives IEEE Signal Process. detection using high-frequency antenna motion and carrier
Mag. 34 27–37 phase data Proc. 26th Int. Technical Meeting Satellite
[4] Jiang C, Chen S, Chen Y, Bo Y, Xia Q and Zhang B 2018 Division Institute Navigation vol 4 pp 2949–91
Analysis of the baseline data-based GPS spoofing detection [21] Jafarnia A, Broumandan A and Nielsen J 2012 GPS
algorithm Monterey, CA, USA Proc. IEEE/ION Position, vulnerability to spoofing threats and a review of
Location Navigat. Symp. (PLANS) pp 397–403 anti-spoofing techniques Int. J. Navig. Obs. 2012 1–16
[5] Wang F, Hu C and Wu S 2020 Research on BeiDou [22] NI S Y, Cui J H, Cheng N P and Liao Y R 2008 Detection and
anti-spoofing technology based on comprehensive radio elimination method for deception jamming based on an
determination satellite service Satell. Navig. 1 1–9 antenna array Int. J. Distrib. Sens. Netw. 14 1–9
[6] Hu Y, Bian S, Li B and Zhou L 2018 A novel array-based [23] Humphreys T, Ledvina B and Psiaki M 2008 Assessing the
spoofing and jamming suppression method for GNSS spoofing threat development of a portable GPS civilian
receiver IEEE Sens. J. 18 2952–8 spoofer Proc. 21st Int. Technical Meeting Satellite Division
[7] Broumandan A, Siddakatte R and Lachapelle G 2017 Feature Institute Navigation pp 2314–25
paper: an approach to detect GNSS spoofing IEEE Aerosp. [24] Broumandan A, Jafarnia A and Dehghanian V 2012 GNSS
Electron. Syst. Mag. 32 64–75 spoofing detection in handheld receivers based on signal
[8] Bardout Y 2011 Authentication of GNSS position: an spatial correlation Proc. 2012 IEEE/ION Position, Location
assessment of spoofing detection methods Proc. 24th Int. and Navigation Symp. pp 479–87
Technical Meeting Satellite Division Institute Navigation [25] Lee J, Kwon K, An D and Shim D 2015 GPS spoofing
(ION GNSS 2011) pp 436–46 detection using accelerometers and performance analysis
[9] Humphreys T 2013 Detection strategy for cryptographic with probability of detection Int. J. Control Autom. Syst.
GNSS anti-spoofing IEEE Trans. Aerosp. Electron. Syst. 13 951–9
49 1073–90 [26] James T and Broumendan A 2017 On the use of low-cost
[10] Wesson K, Rothlisberger M and Humphreys T 2012 Practical IMUs for GNSS spoofing detection in vehicular
cryptographic civil GPS signal authentication Navigation applications Toulouse, France Int. Technical Symp. on
59 177–93 Navigation and Timing (ITSNT) pp 1–8
[11] Kerns A, Wesson K and Humphreys T 2014 A blueprint for [27] Liu Y, Li S, Fu Q, Liu Z and Zhou Q 2019 Analysis of Kalman
civil GPS navigation message authentication Proc. IEEE filter innovation-based GNSS spoofing detection method for
Position, Location Navigat. Symp. (PLANS) pp 262–9 INS/GNSS integrated navigation system IEEE Sens. J.
[12] Hwang P and McGraw G 2014 Receiver Autonomous Signal 19 5167–78
Authentication (RASA) based on clock stability analysis [28] Semanjski S, Muls A and Semanjski I 2019 Use and validation
Proc. IEEE Position, Location Navigat. Symp. (PLANS) of supervised machine learning approach for detection of
pp 270–81. GNSS signal spoofing 2019 Int. Conf. on Localization and
[13] Chen Z, Li H, Broumandan A, Wen J and Lu M 2020 Moving GNSS (ICL-GNSS) pp 1–6
Receiver PVT Solution Authentication Based on [29] Koji I 2018 An examination of GPS spoofing defense with
Monitoring the Combination of Clock Bias and Drift San machine-learning IEICE Tech. Rep. 118 89–94
Diego, California Proceedings of the 2020 International [30] Li J, Zhu X, Ouyang M, Li W, Chen Z and Dai Z 2020
Technical Meeting of The Institute of Navigation Research on multi-peak detection of small delay spoofing
pp 1090–100 signal IEEE Access 8 151777–87
[14] Akos D 2012 Who’s afraid of the spoofer? GPS/GNSS [31] Morales R, Richter P, Falletti E, Fuente A and Lohan E 2020 A
spoofing detection via automatic gain control (AGC) survey on coping with intentional interference in satellite
Navigation 59 281–90 navigation for manned and unmanned aircraft IEEE
[15] Jafarnia A, Broumandan A and Nielsen J 2012 GPS spoofer Commun. Surv. Tutor. 22 249–91
countermeasure effectiveness based on signal strength, [32] Risbud P, Gatsis N and Taha A 2019 Vulnerability analysis of
noise power and C/N0 observables Int. J. Satell. Commun. smart grids to GPS spoofing IEEE Trans. Smart Grid
Netw. 30 181–91 10 3535–48
[16] Psiaki M and Humphreys T 2016 GNSS spoofing and [33] Wylie T R 2013 The Discrete Fréchet Distance and
detection Proc. IEEE 104 1258–70 Application Montana State University Bozeman pp 1–45
[17] Gross J, Kilic C and Humphreys T 2019 Maximum-likelihood [34] Alt H and Godau M 1995 Computing the Fréchet distance
power-distortion monitoring for GNSS signal authentication between two polygonal curves Int. J. Comput. Geom. Appl.
IEEE Trans. Aerosp. Electron. Syst. 55 469–75 5 75–91

11