Università degli Studi di Padova

Department of Information Engineering

Master Thesis in ICT for Internet and Multimedia

Parametric models for a database of realistic threats

to GNSS receivers

Supervisor Master Candidate

Nicola Laurenti Nicola Boscaro
Università di Padova

Academic year: 2019/2020

March 2nd , 2020
ii
To my Parents, my Girlfriend and my Friends.
iv
 Abstract

Threats to GNSS receivers are becoming increasingly complex and easier to implement due
to technological advancement. So, these attacks have become now a serious problem for any
user, not only, for example, for military or safety-of-life purposes anymore. In this context,
Testbed for Attacks and Mititgations has been created to collect data about these attacks and
possible mitigations.
This thesis describes how tested threat scenarios to GNSS signals have been parameterized
to be inserted in the TAM database.

v
vi
 Contents

Abstract v

List of figures ix

List of tables xi

Listing of acronyms 1

1 Introduction 5

2 GNSS technologies 7
2.1 Systems overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 GPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Galileo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.3 GLONASS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.4 BeiDou . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.5 Compatibility and Interoperability . . . . . . . . . . . . . . . . . 12
2.2 GNSS signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 The GPS signal . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.2 The Galileo signal . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 GNSS architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.1 Space segment . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.2 Control segment . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.3 User segment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3 GNSS threats 27
3.1 Jamming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1.1 Jamming attacks . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.2 Jamming detection . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Spoofing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.1 Spoofing attacks . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2.2 Spoofing detection . . . . . . . . . . . . . . . . . . . . . . . . . 36

4 GNSS threat analysis 39

4.1 Testbed for Attacks and Mitigations . . . . . . . . . . . . . . . . . . . . . 39
4.1.1 Equipment setup for the tests . . . . . . . . . . . . . . . . . . . . 40

vii
 4.1.2 Data level spoofing software . . . . . . . . . . . . . . . . . . . . 41
4.2 Threat modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2.1 Jamming modeling . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.2.2 Spoofing modeling . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3 Tests and results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3.1 Jamming tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3.2 Spoofing tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3.3 Jam and Spoof tests . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3.4 Jam and Spoof tests (shorter Jamming duration) . . . . . . . . . . 55

5 Conclusion 57

Appendix A Error sources 59

A.1 Errors from satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
A.2 Errors from receivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
A.3 Errors from signal propagation . . . . . . . . . . . . . . . . . . . . . . . 60

References 61

Acknowledgments 65

viii
 Listing of figures

1.1 GNSS global installed base by segment. . . . . . . . . . . . . . . . . . . . 6

2.1 GPS, Galileo navigational frequency bands. . . . . . . . . . . . . . . . . . 13

2.2 GPS navigation message. . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Modulation scheme for Galileo signals. . . . . . . . . . . . . . . . . . . . 17
2.4 Galileo navigation messages. . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.5 GNSS architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.6 GNSS trilateration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.7 GPS control segment map. . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.8 Antenna gain patterns. . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.9 Correlation function for a Galileo satellite. . . . . . . . . . . . . . . . . . 26

3.1 Typical GNSS vulnerabilities. . . . . . . . . . . . . . . . . . . . . . . . . 28

3.2 Composite jammer classification accounting for both signal and device char-
acteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3 Impact of jamming on a GNSS receiver on AGC count. . . . . . . . . . . . 30
3.4 Comparison of the CAF during acquisition phase with and without a jam-
ming interference. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.5 Positioning results around the true coordinates in a single-frequency jam-
ming test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.6 Jammer parameters organized into trees. . . . . . . . . . . . . . . . . . . 31
3.7 Different approaches for jamming detection. . . . . . . . . . . . . . . . . 33
3.8 GNSS spoofing attack. . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.1 Equipment setup for the tests. . . . . . . . . . . . . . . . . . . . . . . . . 40

4.2 Power computed by the signal analyzer for a GPS signal from one satellite. . 41
4.3 Columns added to the imported csv by EditDF class. . . . . . . . . . . . . 42

ix
x
 Listing of tables

2.1 Overview of GNSS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2 GPS navigation signals. . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Galileo navigation signals. . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.1 Classification of specific jamming attacks. . . . . . . . . . . . . . . . . . . 32

3.2 Classification of specific spoofing attacks. . . . . . . . . . . . . . . . . . . 35
3.3 Highest Risk Navigation Data Parameters. . . . . . . . . . . . . . . . . . 36
3.4 Summary of spoofing detection techniques. . . . . . . . . . . . . . . . . . 38

4.1 Example of the key-value pairs needed to model a navigation scenario. . . . 45

4.2 Example of the key-value pairs needed to model a jamming attack. . . . . . 46
4.3 Example of the key-value pairs needed to model a spoofing attack. . . . . . 47
4.4 Results from the CW jamming attacks. . . . . . . . . . . . . . . . . . . . 48
4.5 Example of the key-value pairs needed to model a CW jamming attack. . . . 48
4.6 Results from the broad band AWGN jamming attacks. . . . . . . . . . . . 48
4.7 Example of the key-value pairs needed to model a broad band AWGN jam-
ming attack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.8 Results from the narrow band AWGN jamming attacks. . . . . . . . . . . 50
4.9 Results from the chirp jamming attacks. . . . . . . . . . . . . . . . . . . . 50
4.10 Example of the key-value pairs needed to model a chirp jamming attack. . . 50
4.11 Results from the frequency hopping jamming attacks. . . . . . . . . . . . 51
4.12 Example of the key-value pairs needed to model a frequency hopping jam-
ming attack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.13 Results from the fixed spoofing at about 100m attacks. . . . . . . . . . . . 52
4.14 Example of the key-value pairs needed to model a fixed spoofing at about
100m attack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.15 Results from the fixed spoofing at about 1km attacks. . . . . . . . . . . . . 53
4.16 Results from the fixed spoofing at about 100km attacks. . . . . . . . . . . 53
4.17 Jam and spoof tests time organization. . . . . . . . . . . . . . . . . . . . 54
4.18 Results from the jam and spoof at about 100m attacks. . . . . . . . . . . . 54
4.19 Results from the jam and spoof at about 1km attacks. . . . . . . . . . . . . 54
4.20 Results from the jam and spoof at about 100km attacks. . . . . . . . . . . 55
4.21 Example of the key-value pairs needed to model a jam and spoof at about
100m attack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.22 Results for the jam and spoof tests with different jamming duration. . . . . 56

xi
xii
 Listing of acronyms

ADC . . . . . . . . . . Analog to Digital Conversion

AGC . . . . . . . . . . . Automatic Gain Control

AR . . . . . . . . . . . . Axial Ratio

AS . . . . . . . . . . . . . Authorized Service

AWGN . . . . . . . . . Additive White Gaussian Noise

BDS . . . . . . . . . . . BeiDou Navigation Satellite System

BGD . . . . . . . . . . . Broadcast Group Delays

BPSK . . . . . . . . . . Binary Phase Shift Keying

CDMA . . . . . . . . Code Division Multiple Access

CNSA . . . . . . . . . China National Space Administration

CRC . . . . . . . . . . . Cyclic Redundancy Check

CS . . . . . . . . . . . . . Commercial Service

CW . . . . . . . . . . . . Continuous Wave

DEI . . . . . . . . . . . . Department of Information Engineering

DLL . . . . . . . . . . . Delay Lock Loop

DoS . . . . . . . . . . . Denial of Service

EGNOS . . . . . . . . European Geostationary Navigation Overlay Service

EC . . . . . . . . . . . . . European Commission

ESA . . . . . . . . . . . European Space Agency

1
EU . . . . . . . . . . . . European Union

FLL . . . . . . . . . . . Frequency Locked Loop

GAGAN . . . . . . . GPS Aided Geo Augmented Navigation

GEO . . . . . . . . . . . GEostationary Orbits

GIS . . . . . . . . . . . . Galileo Initial Services

GLONASS . . . . . GLObal’naja NAvigacionnaja Sputnikovaja Sistema

GNSS . . . . . . . . . . Global Navigation Satellite System

GST . . . . . . . . . . . Galileo System Time

HAS . . . . . . . . . . . High Accuracy Service

ICG . . . . . . . . . . . International Committee on Global Navigation Satellite Systems

IF . . . . . . . . . . . . . Intermediate Frequency

IGSO . . . . . . . . . . Inclined GeoSynchronous Orbits

LHCP . . . . . . . . . Left-Hand Circular Polarization

LO . . . . . . . . . . . . Local Oscillator

MEO . . . . . . . . . . Medium Earth Orbiting

MSAS . . . . . . . . . Multi-functional Satellite Augmentation System

NAVSTAR GPS . NAVigation Signal Timing And Ranging Global Position System

OS . . . . . . . . . . . . Open Service

PLL . . . . . . . . . . . Phase Lock Loop

PNT . . . . . . . . . . . Positioning, Navigation, Timing

PVT . . . . . . . . . . . Positioning, Velocity, Timing

PPS . . . . . . . . . . . . Precise Positioning System

2
PRS . . . . . . . . . . . Public Regulated Service

QZSS . . . . . . . . . . Quasi-Zenith Satellite System

RF . . . . . . . . . . . . Radio Frequency

RFI . . . . . . . . . . . . Radio Frequency Interference

RHCP . . . . . . . . . Right-Hand Circular Polarization

SA . . . . . . . . . . . . . Selective Availability

SaR . . . . . . . . . . . . Search-and-Rescue

SBAS . . . . . . . . . . Satellite-Based Augmentation Systems

SDR . . . . . . . . . . . Software Defined Radio

SIS . . . . . . . . . . . . Signal-in-Space

SISA . . . . . . . . . . . Signal-in-Space Accuracy

SMS . . . . . . . . . . . Short Message Service

SPS . . . . . . . . . . . . Standard Positioning System

SQM . . . . . . . . . . Signal Quality Monitoring

SV . . . . . . . . . . . . . Space Vehicle

SVID . . . . . . . . . . Space Vehicle IDentifiers

TAM . . . . . . . . . . Testbed for Attacks and Mitigations

TOA . . . . . . . . . . . Time of Arrival

TOW . . . . . . . . . . Time-of-Week

USSR . . . . . . . . . . Union of Soviet Socialist Republics

UTC . . . . . . . . . . . Coordinated Universal Time

WAAS . . . . . . . . . Wide Area Augmentation System

WADS . . . . . . . . . Wide Area Differential Service

3
4
 Introduction
1
A Global Navigation Satellite System (GNSS) is an infrastructure that allows users
with a compatible device to determine their position, velocity and time by processing signals
from satellites. Altough it was born for military purposes decades ago, nowadays, it is used
globally for a wide range of services, as we can see in Fig. 1.1, to the point where its absence
would be inconceivable.
Right now there is a digital revolution that is influencing our lives, but also the business
models of companies that are constantly trying to keep up with “macrotrends” (i.e. climate
change and sharing economy just to name a few) in technological development. Indeed,
when these services require positioning and timing data, the adoption of GNSS solutions is
needed, so that we may have environmentally friendly transport solutions, sustainable agri-
culture, meteorology and climate change monitoring or ride sharing services.
Of course, with more and more people that are growing concerns about security over the
years, GNSS threats may now be considered hot topic given the spreading of GNSS services.
The low power of the signals that reach the Earth surface from satellites, in fact, makes it easier
to perform denial of service (jamming) or forging of GNSS signal (spoofing) attacks. The
impact of such threats may be so disruptive - just imagine businesses that base their activities
on GNSS, or safety of life services - that the study of countermeasures to these problems is
truly significant.
This thesis was developed as part of the project Testbed for Attacks and Mitigations (TAM)

5
 (a) Global installed base by segment

(b) Installed base of ”Other” segments

Figure 1.1: GNSS global installed base by segment (1.1a) and detailed view of ”Other” segment (1.1b). [1]

during an internship carried out between August 2019 and February 2020, and it is about the
laboratory testing of several threat scenarios to GPS and Galileo signals and the following
insertion, based on specific parameters, of the results in a database needed for the services
offered by TAM. This thesis is composed as follows:

Chapter 2 a brief overview of GNSS technologies with regards to different systems services and
architecture, and the description of the GNSS signal.

Chapter 3 a description of GNSS threats jamming and spoofing, with some more details about
some of the most common attacks and their detection.

Chapter 4 a report about the results achieved from the tests performed during the internship
with also a description of what has been done in those months.

Conclusion conclusions from the work carried out and suggestions concerning possible future
research topics.

Appendix few more details about GNSS error sources.

6
 GNSS technologies
2
Satellite navigation systems have history that starts during what is now called Space
Race* , starting with the first system deployed in the 1960s, Transit by the US military, which
was based on the Doppler effect. To achieve enough accuracy this system needed constant
monitoring from the base station which would send information about the deviation in satel-
lites orbit so that next satellite broadcast would have had updated ephemeris.
What we have, about 60 years later in the modern GNSS, (e.g. such as GPS or Galileo
just to name a few), is a more direct situation, where the satellite broadcasted signals directly
contains orbital data and the precise time the signal was transmitted.
The description in this chapter is based on [3], [4], [5] and [6] and is not intended to be
comprehensive, but only the basic concepts of modern GNSS and everything that will be
needed in the following chapters to better understand this thesis work.

2.1 Systems overview

Global Navigation Satellite System (GNSS) is the name used to refer to a constellation of
satellites, usually composed of 18-30 units in medium Earth orbit (MEO) spread between
several orbital planes, providing signals from space that transmit positioning and timing data
* The competition between URSS and US, that began on August 2, 1955 during the Cold War, to achieve
firsts in spaceflight capability. [2]

7
 System GPS Galileo GLONASS BeiDou
MEO,
Orbit MEO MEO MEO
IGSO, GEO
Nominal number
24 30 24 27, 3, 5
of satellites
6 planes, Walker (24/3/1), Walker (24/3/1), Walker (24/3/1),
Constellation
56° inclination 56° inclination 64.8° inclination 55° inclination
OS, HAS, PRS, OS, AS, WADS,
Services SPS, PPS SPS, PPS
SaR SMS
Initial service Dec 1993 Dec 2016 Sep 1993 Dec 2012
Origin USA Europe Russia China
E1 1575.42
L1 1575.42 L1 1602.00 B1 1561.098
Frequency E5a 1176.45
L2 1227.60 L2 1246.00 B2 1207.14
(MHz) E5b 1207.14
L5 1176.45 L3 1202.025 B3 1268.52
E6 1278.75

SPS: Standard Positioning Service; PPS: Precise Positioning Service; OS: Open Service; HAS:
High Accuracy Service; PRS: Public Regulated Service; SaR: Search-and-Rescue service; AS:
Authorized Service; WADS: Wide Area Differential Service; SMS: Short Message Service
Galileo services and initial service date were changed based on [7] and [8]
Table 2.1: Overview of GNSS. [6]

to GNSS receivers. The actual systems vary, but use orbital inclinations of about 55° and have
orbits, at an altitude of about 20,000 kilometres, with periods of roughly twelve hours (see
Table 2.1).

Currently there are four active navigation satellite systems that provide global coverage
which are the American GPS, the European Galileo, the Russian GLONASS and the Chinese
BeiDou. Since a GNSS signal has to satisfy four performance parameters (accuracy, integrity,
continuity and availability), several regional Satellite-Based Augmentation Systems (SBAS)
come to help GNSS by correcting signal measurement errors (see Appendix A for different
GNSS error sources). Some examples are the American Wide Area Augmentation System
(WAAS), the European Geostationary Navigation Overlay Service (EGNOS), the Japanese
Multi-functional Satellite Augmentation System (MSAS) or Quasi-Zenith Satellite System
(QZSS), and the Indian GPS Aided Geo Augmented Navigation (GAGAN).

8
2.1.1 GPS

The Global Positioning System (GPS) is the American GNSS, implemented and operated
by the US Department of Defense (DoD), which was declared fully operational from April
1995 with the baseline GPS being specified for 24 satellites, even if, currently [9], the system
employs 31 Block IIA/IIF/IIR/IIR-M satellites† .
The GPS, originally NAVSTAR GPS, is a line-of-sight, all weather, world-wide continu-
ously available satellite-based RF positioning system that provides 3-dimensional position,
velocity and time data to an end-user with an appropriate receiver. There are two differ-
ent services provided by the GPS: the Standard Positioning Service (SPS) and the Precise
Positioning Service (PPS). SPS is the service available for the civilian use which use the pub-
lic C/A code on the L1 carrier. On the other hand, PPS is the service used by the military,
and other authorized parties, which uses the encrypted P(Y) codes on both L1 and L2 car-
riers. Anyway, after the deactivation of the Selective Availability (SA) technique on 2 May
2000 by the administration Clinton, which was done in order to make GPS more respon-
sive to civil and commercial users world-wide [11], the signals that the civilian receives are
not so degraded anymore and the differences between the two services are that the PPS also
features increased robustness, higher resistance to jamming, improved accuracy regarding sig-
nal distortion caused by ionosphere and multipath propagation. It must be said anyway that
nowadays, even if the encrypted code P(Y), which will be replaced by the new more power-
ful M-code, is still available only for military use, there exists dual-frequency GPS equipment
which is available for civilian use, that significantly improves the obtained results, but its cost
and size has limited it to professional applications.
The modern GPS receivers also have the possibility to use an extra civilian safety-of-life sig-
nal on L5 band which can have much higher accuracy, pinpointing to within 30 centimetres.
[12]

2.1.2 Galileo

The European GNSS Galileo is, as opposed to GPS, GLONASS or BeiDou, under civilian
control. This means that, while the formers may be switched off or made less precise when
desired (e.g. in case of conflict [13]) causing a very costly disruption, this situation should not
happen for the Galileo system.

†
From 22 August 2019, 2 third generation GPS Block IIIA satellites are also in orbit [10]

9
 In the 1990s the European Union (EU) saw the need for Europe to have its own GNSS
so, together with private investors, they started the Galileo project for a civilian GNSS. This
project had many problems over the years, such as the nationalisation of the system in 2006
due to the falling apart of the public/private partnership and the subsequent funding prob-
lems, but in the end European Commission (EC) together with the European Space Agency
(ESA) made it and the European GNSS went live in the end of 2016.
The Galileo space segment has right now 26 satellites in space, of which 5 are not usable for
the services offered by this GNSS [14], but will include, once fully deployed, a constellation
of 30 MEO satellites (including 6 spares).
The Galileo Initial Services (GIS) which are currently available and offered are the Open
Service (OS), the Public Regulated Service (PRS) and the Search-and-Rescue service (SaR)
[7]. OS is the freely accessible service that targets the mass market and is intended for mo-
tor vehicle navigation and location-based mobile telephone services in high-volume satellite
radio navigation applications scenarios. It requires no authentication and it is expected that
this service is used along with GPS to improve performance in severe environments, or with
GNSS Augmentation techniques if one needs to achieve higher precision without the ne-
cessity of integrity. For sensitive applications there is PRS which, like PPS for the GPS, is
restricted to government or otherwise authorised users. The last service of GIS is SaR, a ser-
vice that is the European contribution to the international COSPAS-SARSAT co-operative
effort on humanitarian search and rescue activities. SaR helps to forward distress signals to a
rescue coordination centre by detecting emergency signals transmitted by beacons and relay-
ing messages to them. Moreover, since for Galileo system there is always at least one satellite
which is in view of any point on Earth, we can achieve near real-time distress alerts.
Once the satellite constellation and the ground infrastructure will be completed, another
service will also be available, the High Accuracy Service (HAS)‡ . The HAS is complementary
to OS and it provides an additional navigation signal and added value services in a different
frequency band, allowing the user to obtain an accuracy to the nearest centimetre. Like OS,
this service does not offer integrity information, but, in addition to high accuracy, it can offer
the authentication of the information encoded in the signal and the signal Time of Arrival
(TOA)§ . [8]

‡
It is a rescope of the former Galileo Commercial Service (CS).
§
While integrity protection protects a receiver from, for example, satellite failures (orbit or clock), this one
provides a defense against threats, such as jamming and spoofing, due to the use of encryption.

10
2.1.3 GLONASS

The former Soviet Union developed in the 1980s the GLObal’naja NAvigacionnaja Sputniko-
vaja Sistema (GLONASS), its own GNSS operated by Russian military but also available for
civilian use, with the first test satellite launched on 12 October 1982. GLONASS was then
finally deployed with full operational capability (24 satellites) in 1995. However, due to the
dismantling of the USSR, by 2002 the constellation had dropped to as few as seven satellites.
It is because of the support by the Russian government that GLONASS was reborn and
reached full operational capability again on 8 December 2011.
The services offered by GLONASS are the same given by the GPS, SPS and PPS.

2.1.4 BeiDou

Started as a regional navigation satellite system with a program of research and development
that began in 1980, also the Chinese BeiDou Navigation Satellite System (BDS, formerly
known as COMPASS) is now a GNSS operated by the Chinese Spacial Agency CNSA.
There were 3 stages in the development of this GNSS that has brought us to the present
BeiDou. The first, which was from 2000 to 2006, was the BeiDou-1 and it was a regional
navigation satellite system. It consisted of 4 satellites in geostationary orbit (GEO), which
means that the system did not need a large constellation of satellites, but obviously it limited
the coverage to areas on Earth where the satellites were visible. From 2007 to 2019, the sec-
ond version of BDS, the BeiDou-2, which planned to have a constellation of 35 satellites: 5 in
GEO for backward compatibility with BDS-1, 27 in MEO and 3 in inclined geosynchronous
orbit (IGSO), was the first Chinese test for a GNSS. It reached the full operational capability
at the end of 2012 with 16 satellites in orbit of which 14 were in service and it covered China
and its surrounding regions. To achieve full global coverage, from 2015 CNSA started the
new GNSS project, BeiDou-3, and planned for it to include 5 GEO satellites, 3 IGSO satel-
lites and 24 MEO satellites. It is expected to become operational by 2020 and it will offer
2 global services, Open Service (OS, similar to the GPS one) and Authorized Service (AS),
and 2 regional services, Wide Area Differential Services (WADS) and Short Message Service
(SMS).
Currently there are in space: no operational BDS-1 satellites, 15 operational BDS-2 satellites
and 28 BDS-3 satellites of which 19 operational and 9 in testing. [15]

11
2.1.5 Compatibility and Interoperability

As we have just seen, there exist more global navigation satellite systems and, without taking
into consideration possible political reasons behind this situation, the technical cause for it
is that a single GNSS may not be enough to guarantee user performances. This leads us to
two main problems [16]:

• Compatibility: different systems should not interfere with each other.

• Interoperability: we would like to use these systems, that are independent by design,
together to provide better capabilities at user level.

The United Nations’ International Committee on Global Navigation Satellite Systems

(ICG), which promotes voluntary cooperation on matters of mutual interest related to civil
satellite-based positioning, navigation, timing (PNT) and value-added services, has encour-
aged coordination among providers of GNSS, but also regional systems and augmentations,
about these topics for years.
The compatibility issue was addressed, specifically between GPS and Galileo but then also
for other systems, by the EU-US Agreement on the Promotion, Provision and Use of Galileo
and GPS Satellite-Based Navigation Systems and Related Applications in 2004 [17]. This
document has set up the models and methodology for the radio frequency compatibility of
satellite navigation systems.
As regards interoperability, Galileo was designed to be interoperable with GPS and Bei-
Dou achieved it in 2017. As regards GLONASS, even if a working group about this topic
was established in 2004 and had a “successful meeting” in 2006, everything about this col-
laboration seems on hold from April 2014. It should be said that it is anyway possible to
use a receiver that uses GLONASS and other GNSS constellations to achieve better results.
This is because the Russian GNSS is considered system interoperable, so, from GPS and
GLONASS, we will obtain the same measurements within the specified accuracy of each
individual system, and those may be used together to achieve higher accuracy. [18]

2.2 GNSS signals

GNSS satellites continuously transmit navigation signals at two or more frequencies in L

band. These signals contain ranging codes and navigation data to allow users to compute
both the travel time from the satellite to the receiver and the satellite coordinates at any epoch.
The main signal components are described as follows:

12
Figure 2.1: GPS, Galileo navigational frequency bands. [20]

• Carrier: radio frequency sinusoidal signal at a given frequency fRF .

• Ranging code, C(t): sequences of zeros and ones which allow the receiver to determine
the travel time of the radio signal from the satellite to the receiver. They are called PRN
sequences or PRN codes.

• Navigation data, D(t): a binary-coded message providing information on the satellite

ephemeris (pseudo-Keplerian elements or satellite position and velocity), clock bias pa-
rameters, almanac (with a reduced-accuracy ephemeris data set), satellite health status
and other complementary information.

GNSS signals are usually transmitted with quadrature modulation using Right-Hand Cir-
cular Polarization (RHCP), and the overall transmitted signal, the Signal-in-Space (SIS), can
be expressed as:
√
s(t) = 2P C(t)D(t) cos(2πfRF t + φ0 ), (2.1)

where P is the average power of the sinusoidal signal and φ0 is the initial phase. C and D
have an amplitude of ±1 varying with time. [19]
In the next sections I will focus on GPS and Galileo signals description, since those are the
system with which I have made the tests for this thesis.

2.2.1 The GPS signal

Legacy GPS signals are transmitted on two radio frequencies, L1 and L2, and their frequen-
cies are derived from a fundamental frequency, f0 = 10.23MHz , generated by onboard

13
atomic clocks:

• fL1 = 154 × 10.23MHz = 1575.420MHz

• fL2 = 120 × 10.23MHz = 1227.600MHz

where L1 is used by SPS while PPS is a dual frequency service that uses both L1 and L2.
To send different signals on the same radio frequency, GPS uses the Code Division Mul-
tiple Access (CDMA) technique and the Binary Phase Shift Keying (BPSK) as modulation
method.
The following types of PRN codes and messages are modulated over the two carriers

• Coarse/Acquisition (C/A) code, also known as civilian code, C(t): this sequence con-
tains 1023 bits and is repeated every millisecond (i.e. a chipping rate of 1.023 Mbps).
Then, the duration of each C/A code chip is 1 µs, which means a chip width or wave-
length of 293.1 m. This code is modulated only on L1. The C/A code defines the SPS.

• Precision code, P (t): This is reserved for military use and authorised civilian users.
The sequence is repeated every 266 days (38 weeks) and a weekly portion of this code
is assigned to every satellite, called the PRN sequence. Its chipping rate is 10 Mbps,
which leads to a wavelength of 29.31 m. It is modulated over both carriers L1 and L2.
This code defines the PPS.
In order to protect military receivers against an adversary transmitting a faulty copy
of the GPS signal to mislead the receiver, and to deny access of non-authorized users
to the precise ranging code P , the latter is encrypted by combining it with a secret W
code (called security code), resulting in the Y code, which is modulated over the two
carriers L1 and L2.

So, the SIS transmitted by a GPS satellite takes the following form:

s(t) = sL1 (t) + sL2 (t), (2.2)

where
√
sL1 (t) = 2PP,1 W (t)C(t)D(t) sin(2πfL1 t + φL1 )+
√
+ 2PC C(t)D(t) cos(2πfL1 t + φL1 ),
√
sL2 (t) = 2PP,2 W (t)P (t)D(t) sin(2πfL2 t + φL2 ).

14
 Carrier freq. Code rate Data rate
Link PRN code Modulation type Service
(MHz) (Mcps) (bps)
C/A BPSK(1) 1.023 50 Civil
P BPSK(10) 10.23 50 Military
L1 1575.420 M BOCsin(10,5) 5.115 N/A Military
L1C-I data 50
MBOC(6,1,1/11) 1.023 Civil
L1C-Q pilot -
P BPSK(10) 10.23 50 Military
M 25
L2 1227.600 L2C BPSK(1) Civil
L 1.023 -
M BOCsin(10,5) N/A Military
L5-I data 50
L5 1176.450 BPSK(10) 10.23 Civil
L5-Q pilot -
Table 2.2: GPS navigation signals. [4]

Recently there was a GPS modernisation which brought to us an additional frequency

L5 (fL5 = 115 × 10.23MHz = 1176.450MHz ), for life critical applications, and several
new ranging codes on the different carrier frequencies (civilian signals L2C, L5C, L1C and the
military M-code, which have different modulation methods). For more details about GPS
signals and frequencies see Table 2.2.
Since a receiver needs to know the time and position of each active satellite, GPS encodes
this information into the navigation message and modulates it over both carriers at 50 bps.
In particular, what are contained in the navigation message are 25 frames of 30 s each, for a
total of 12.5 min of transmission, which in turn is subdivided in five sub-frames of 6 s each.
Every sub-frames, then, consists of 10 words with 30 bits per word.
Every sub-frame always starts with the telemetry word TLM, for synchronisation, and
the transference word (HOW), which provides the seconds of the GPS week that allow the
receiver to acquire the week-long P (Y ) code segment. Except for these fixed elements, the
contents of every sub-frame are as follows:

• Sub-frame 1 – Satellite Clock and Health Data: contains, first of all, clock information,
which is needed to compute at what time the navigation message is transmitted from
the satellite. Additionally, this sub-frame also contains health data indicating whether
or not the data should be trusted.
• Sub-frames 2 and 3 – Satellite Ephemeris Data: contain the satellite ephemeris data.
The ephemeris data relate to the satellite orbit and are needed to compute a satellite
position.

15
Figure 2.2: GPS navigation message. [3]

• Sub-frames 4 and 5 – Support Data: the last two sub-frames contain almanac data,
which are the ephemerides and clock data with reduced precision. Additionally, each
satellite transmits almanac data for all GPS satellites while it only transmits ephemeris
data for itself. The remainder of sub-frames 4 and 5 contain various data (e.g. UTC
parameters, health indicators, and ionospheric parameters).

2.2.2 The Galileo signal

The Galileo system transmits three signals: E1, E5 (consisting of E5a and E5b) and E6:

• fE1 = 1575.420MHz

• fE5a = 1176.450MHz

• fE5b = 1207.140MHz

• fE6 = 1278.750MHz

where E1 supports OS and PRS, E5a and E5b only OS, and E6 is used for PRS. When HAS
will be available, it will be supported by E1, E5b and E6.

16
(a) E1 CBOC (b) E5

(c) E6

Figure 2.3: Modulation scheme for Galileo signals: E1 CBOC in 2.3a, E5 in 2.3b and E6 in 2.3c [20].

As regards Galileo SIS, it is comparable to GPS one except for the fact that we have more
signals¶ . In Fig. 2.3 you can see more details about Galileo signals.
As in GPS, all satellites share the same frequencies and the signals are differentiated by the
CDMA technique. Moreover, some signals can contain data and pilot channels but, while
they both provide ranging codes, only the data channel also includes navigation data. So, the
pilot channel, which is considered a data-less channel, is needed to help tracking weak signals.
For more details see Table 2.3.
Regarding Galileo navigation message, the Galileo SIS data channels transmit different
message types according to this subdivision:
• F/NAV message type: for the OS on E5a-I.
• I/NAV message type: for the OS on E5b-I and E1-B‖ .

As we can see in Fig. 2.4, we have differences in the navigation message size based on the
message type.
¶
Galileo signals are more like the modernised version of GPS signals instead of the original ones.
‖
E1-B will be also for the HAS in the future

17
 Carrier freq. Channel or Code rate Data rate
Band Modulation type Service
(MHz) sig. comp. (Mcps) (bps)
E1-A data BOCcos(15,2.5) 2.5575 N/A PRS
E1 1575.420 E1-B data 125
MBOC(6,1,1/11) 1.023 OS, HAS
E1-C pilot -
E5a-I data 25
E5a 1176.450 BPSK(10) 10.23 OS
E5a-Q pilot -
E5b-I data 125
E5b 1207.140 BPSK(10) 10.23 OS
E5b-Q pilot -
E6-A data BOCcos(10,5) N/A PRS
E6 1278.140 E6-B data 5.115 500
BPSK(10) HAS
E6-C pilot -
Table 2.3: Galileo navigation signals. [4]

In the F/NAV message, every sub-frame always starts with the page type field (6 bits),
which is needed to identify the page content, then there is the navigation data field (208
bits) and a CRC (24 bits). In particular, each sub-frame is composed as follows:

• Page 1: contains SVID, clock correction, SISA, ionospheric correction, BGD, Signal
health status, GST and Data validity status.

• Page 2, 3 and 4: contain ephemeris, GST, GST-UTC conversion, GST-GPS conver-

sion and TOW.

• Page 5 and 6 : contain almanac of one satellite and part of the almanac of another one.

As regard I/NAV instead, we have two types of pages:

• Nominal pages: having a duration of 2 seconds, they are transmitted sequentially in

time in two parts of duration 1 second each on each of the E5b-I and E1-B components.

• Alert pages: having a duration of 1 second, they are transmitted in two parts of du-
ration 1 second each at the same epoch over the E5b-I and E1-B components. This
transmission is repeated at the next epoch but switching the two parts between the
components.

The content of each sub-frame, even if in a different order, is the same of the F/NAV mes-
sage type, clearly with different layout given that the service provided on these frequencies is
a dual-frequency service.

18
(a) F/NAV (b) I/NAV

Figure 2.4: Galileo navigation messages: F/NAV in 2.4a, I/NAV in 2.4b. [20]

2.3 GNSS architecture

A GNSS basically consists of three main segments (see Fig. 2.5):

• the space segment, which comprises the satellites;

• the control (or ground) segment, which is responsible for the proper operation of the
system;

• the user segment, which includes the GNSS receivers providing positioning, velocity
and precise timing to users.

In the following paragraphs there will be a brief description of the first two segments, but
then I will focus more on the receiver part for the remaining part of the section.

2.3.1 Space segment

The main functions of the space segment are to generate and transmit code and carrier phase
signals, and to store and broadcast the navigation message uploaded by the control segment.
These transmissions are controlled by highly stable atomic clocks onboard the satellites.
The GNSS space segments are formed by satellite constellations with enough satellites to
ensure that users will have at least four satellites in view simultaneously from any point on
Earth’s surface at any time (see Fig. 2.6).

19
Figure 2.5: GNSS architecture. [21]

Figure 2.6: GNSS trilateration. [22]

20
Figure 2.7: GPS control segment map. [23]

2.3.2 Control segment

The control segment (also referred to as the ground segment) is responsible for the proper
operation of the GNSS. Its basic functions are:
• to control and maintain the status and configuration of the satellite constellation;
• to predict ephemeris and satellite clock evolution;
• to keep the corresponding GNSS time scale (through atomic clocks);
• to update the navigation messages for all the satellites.

Since the control performed by the ground segment is essential to ensure that the informa-
tion obtained from the users’ receivers are correct, an extensive distribution of monitoring
and control stations, and ground antennas are needed to always reach each satellite. See in
Fig. 2.7 an example for the GPS control segment.

2.3.3 User segment

The user segment is composed of GNSS receivers. Their main function is to receive GNSS
signals, determine pseudoranges (and other observables) and solve the navigation equations
in order to obtain the coordinates and provide a very accurate time.
A conventional GNSS hardware receiver is basically composed of the analog part, the dig-
ital part including the application processor, and the interfaces for input-output.

21
Antenna

The basic purpose of a GNSS user antenna, operating in the L-band of the radio frequency
spectrum, is the reception of navigation signals from all visible GNSS satellites. Anyway,
this is not a simple task since received GNSS signals from satellites are notoriously weak, and
also they can arrive from, virtually, any direction with signals from different satellites arriving
simultaneously.
Some parameters that defines, or affects, GNSS antennas functionality are:

• Frequency coverage: as we have seen in Table 2.1, there are a lot of frequency bands
available for satellite navigation purposes, so the antenna may need to cover some or
all of these bands. This, together with size requirements, makes the antenna harder to
design.

• Gain pattern: the gain is the ratio of the power delivered by the antenna in response to
a signal arriving from a given direction compared to that delivered by a hypothetical
isotropic reference antenna. The spatial variation of an antenna’s gain is referred to
as the radiation pattern and, under the antenna reciprocity theorem and by ignoring
losses, this is equal to the gain pattern. In a “theoretical world” the antenna should
cover the entire hemisphere above it with no variation in gain (see Fig. 2.8a). But,
since in the “real world” we also have to consider multipath rejection and antenna
noise temperature, we have some roll-off (see Fig. 2.8b).

• Multipath suppression: except for the GNSS signals that arrives directly from the satel-
lite to the receiver’s antenna, they may also be reflected off the ground, buildings, or
other obstacles, and arrive at the antenna multiple times and delayed in time, therefore,
degrading positioning accuracy. This is what is called multipath and the problem in-
troduced by it is due to the fact that reflected signals typically contain a large LHCP
component. Usually LHCP reflections that arrive at the antenna at high elevation an-
gles are not a problem because the axial ratio (AR)** tends to be quite good at these
elevation angles and, therefore, the reflection will be suppressed. On the other hand,
at lower elevation angles, to improve the results it makes sense to have some level of
gain roll-off towards these elevation angles, even if a good AR is also required.

From equation 2.1, the received signal for a visible satellite at the end of a receiver antenna
can be modelled as:
√
rRF (t) = a 2P D(t − tp )C(t − τ ) cos[2π(fRF + fD )(t − tp ) + φ0 ] + nRF (t), (2.3)
** Axial ratio is the measure of the polarization ellipticity of an antenna designed to receive circularly polarized
signals. It is best when AR is close to 1, or 0, dB.

22
(a) Theoretical antenna (b) ”real world” antenna

Figure 2.8: Antenna gain (in dBi) patterns. [24]

with
fRF dtp
fD = −,
c dt
where a is the path attenuation, tp is the propagation time, τ is the propagation time modulo
the code period, denoted as code delay, fD is the carrier Doppler frequency shift (Hz) and
nRF (t) is the additive noise component at RF.

Front-end

The GNSS signal obtained by the receiver’s antenna is then given to the front end. This
part is then responsible for the first handling of the received signals for the following signal
processing tasks. The different steps performed at the receiver’s front-end are the following:

• Filtering and amplification: since the received GNSS signal has low power, after the
antenna there is a set of filtering and low-noise amplification stages. These are needed
to amplify the signal, to compensate for transmission losses, while keeping the noise
figure low and rejecting possible out-of-band interference.

• Down-conversion: this stage’s main objective is to convert the input signal from RF
to IF and/or to baseband. This is achieved through the local oscillator (LO) which
deals with signal mixing operations. These operations are the mixing of two different
frequency signals in order to shift the same information at two different frequencies,
where one is the sum of the two frequencies mixed, while the other is their difference.

• Automatic Gain Control (AGC): in this stage, the AGC is responsible for adjusting the
gain of the front-end section in order to take benefit from the full dynamic range. The
most common implementation is to adjust the signal gain depending on incoming
signal levels, for example by estimating the noise standard deviation.

23
 • Quantization: after the down-conversion, the incoming signals are digitized through
analog to digital converters (ADC), ensuring that quantization errors and dynamic
ranges are appropriate to accommodate the signal’s characteristics.

From the signal received by the antenna, the signal at the end of the front-end, for a single
satellite, can be modelled as

rIF (t) = rIF (k; τ ; φ; fD ; A) =AD(kTs − tp )C(kTs − τ ) cos[2π(fIF + fD )kTs + φ]

+ nIF (t) for k = 0, 1, 2, ...,
(2.4)

where Ts is the sampling time interval (s) such that t = kTs and nIF is the corresponding
noise at IF.

Obtaining observables

After having down-converted and digitized a GNSS signal we can obtain, from the baseband
processing block, observables, that are code pseudo-ranges and carrier phase measurements,
and navigation data.
The digital processing of a GNSS signal in a channel starts with the acquisition, which
is the detection that a signal is present. During this time frame coarse estimates of the code
delay and Doppler of the signal are determined in feed-forward manner, and then the channel
switches to code and carrier tracking to refine the estimates in a feedback structure.

Correlator method In order to detect and track the GNSS signals, the receiver em-
ploys the auto-correlation principle. It generates a transmitted GNSS signal copy of a single
satellite inside the receiver and correlates this replica signal with the received signal. If the
signal parameters in terms of code phase and Doppler shift match reasonably well, then the
correlation value increases. The correlation is realized as an integration of the product of
received and replica signal.
The replica signal can be modelled as

ˆ
r̂IF (t) = r̂IF (k; τ̂ ; φ̂; fˆD ; Â) = 2C(kTs − τ̂ )ej[2π(fIF +fD )kTs +φ̂] (2.5)

and, by replacing Θ = 2π(fIF + fD )kTs + φ and Θ̂ = 2π(fIF + fˆD )kTs + φ̂, we have

24
that the correlation function is

∑
M
corr [rIF (k), r̂IF (k)] = r(k)r̂(k) (2.6)
k=1

where M is the number of samples within the integration time T = M Ts , which is usually
shorter or equal to the navigation data bit/symbol period. From [6] we obtain that

corr [rIF (k), r̂IF (k)] = ĀDR(δτ )sinc(δfD T )ejδφ + η =

(2.7)
= I + jQ

where  is the amplitude of the baseband signal component assuming a normalized noise
component, η is the noise after correlation operation, R(δτ ) is the normalized correlation
function of C(kTs ), δτ is the code delay error (s), δφ is the carrier-phase error (rad) and δfD
is the Doppler error (Hz). Finally, I and Q represent post-correlation values in in-phase and
quadrature components respectively.

I = ĀDR(δτ )sinc(δfD T ) cos(δφ) + ηI

Q = ĀDR(δτ )sinc(δfD T ) sin(δφ) + ηQ

Acquisition The purpose of acquisition is to determine visible satellites and coarse val-
ues of carrier frequency and code-phase of the satellite signals. These values will be then
refined during the tracking. During search and acquisition stage a correlation function is
computed and compared to a specific threshold, then, if the correlation value exceeds that
level, the signal is declared to be present and the position of the peak are the coarse estimates
(see Fig. 2.9).

Tracking After acquisition stage is completed, the GNSS receiver start the tracking stage,
which is characterized by tracking loops that are designated to adjust the input of the lo-
cal replica signal generators to match the received signals. There are three different track-
ing loops architectures: Delay Lock Loop (DLL), for code delay tracking; Phase Lock Loop
(PLL), for carrier-phase tracking; finally the Frequency Locked Loop (FLL), for carrier Doppler
frequency shift tracking.
During tracking the channel synchronizes to the broadcast navigation data message and de-

25
Figure 2.9: Correlation function for a Galileo satellite at low elevation. [6]

codes it. The decoded bit information contains the satellite’s ephemeris and almanac, system
time information, and meteorological parameters. The information from code and carrier
tracking blocks together with time synchronization information is used to generate the pri-
mary measurements of GNSS. Using these, finally, the navigation unit computes the GNSS
navigational equation to obtain the user position, velocity and timing (PVT) information.

26
 GNSS threats
3
GNSS are increasingly being relied upon for safety-related and commercially sensi-
tive applications, such as autonomous vehicles and time-synchronization, so these systems
have become an attractive target for illicit exploitation by different types of attackers for dif-
ferent reasons. Given also that the GNSS signals’ power is very low when they reach Earth sur-
face (about -160 dBW), even a low-power interference can easily jam or spoof GNSS receivers
within a radius of several kilometres. This ease of attack, together with the technological ad-
vance, implies that the number of threats to the different systems is constantly increasing and
they are becoming more sophisticated.
The range of threats and vulnerabilities that can impact a navigation satellite system is
rather wide (see Fig. 3.1) and they can be mainly classified into intentional or unintentional.
What we will focus on in this thesis are two intentional threats which are jamming, a Denial
of Service (DoS) attack, and spoofing, which consists of sending forged GNSS signals to the
receiver.

3.1 Jamming

Jamming is probably the easiest type of attack one can perform on GNSS signals, as it does
not need previous knowledge about them. The classic jamming attack is a form of RFI gener-
ated by devices, called jammers (see Fig. 3.2), which deliberately transmit powerful signals at

27
Figure 3.1: Typical GNSS vulnerabilities. [25]

the GNSS frequencies. The interference signals can be, for example, continuous wave, wide-
band or narrow-band radio frequency signals, or chirp signals, and the higher power has the
jamming signal, the more damage will be caused and the further it will reach.
• Continuous wave: is a signal with constant amplitude A and frequency fCW , typically
a sine. So a continuous wave jamming signal q(t) can be modeled as

q(t) = A sin(2πfCW t).

• Broad or narrow band AWGN : AWGN is generated by a series of Gaussian indepen-

dent and identically distributed variables Xi , with uniform power σ 2 across the whole
frequency band, and zero mean. So each variable can be modeled as

Xi ∼ N (0, σ 2 ), Xi ⊥ Xj ∀i ̸= j.

• Chirp signal: chirp signal sweeps the frequency from low to high frequency linearly.
It can be modeled as a sinusoidal wave that increases in frequency linearly over time

q(t) = A cos(2π(f0 + αt)t),

where f0 is the starting frequency, and α is the frequency variation rate.

28
Figure 3.2: Composite jammer classification accounting for both signal and device characteristics. [26]

With this type of attack, the recovery time for a receiver may be of a few seconds, or several
minutes (based also on the receiver quality), after the interference is eliminated. Depending
on the receiver part that is affected by the jamming threat, usually there could be three dif-
ferent consequences:
• Front-end: in the presence of strong jamming signals several elements of the front-end
(filters, amplifiers) may be led to work outside their nominal regions, generating, for
example, clipping phenomena (signal amplitude exceeding the hardware capability to
treat them). In Fig. 3.3 we can see the impact of jamming on AGC count, which is
significantly reduced by the threat.
• Acquisition: as said in 2.3.3, the main operation performed by the acquisition block
is to correlate input signal with local replicas of the signal code and carrier. In this
respect, the cross-ambiguity function (CAF) is evaluated, in which, when the GNSS
signal is present and in the absence of interference, a single dominant peak should
appear. The peak reveals the signal presence and it is located at the approximate signal
code delay and Doppler shift. As we can see in Fig. 3.4, in the presence of a jamming
attack the peak-to-noise-floor separation decreases as the interfering power increases.
• Tracking: since this stage is in charge of generating GNSS measurements such as pseu-
doranges, Doppler, carrier phase and the amplitude of the received signals, what a
jamming attack causes is a quality deterioration of these measurements. In Fig. 3.5 we
can see the shifted results, caused by a jammer on a u-blox receiver, with respect to the
true coordinates.

29
Figure 3.3: Impact of jamming on a GNSS receiver on AGC count. [26]

Figure 3.4: Comparison of the CAF for a GPS L1 C/A acquisition search space in an interference-free environment
(left) and in the presence of an in-band CW signal at -130 dBW (right). [26]

In the following sections, and in the next chapter, I will use many times the carrier-to-
noise density power ratio C/N0 , that is the ratio of the signal power P and noise PSD N0 .
The C/N0 is continuously estimated by the receiver and it is usually provided in logarithmic
units, dB-Hz. It is used to characterize the relation between signal and jamming powers.

3.1.1 Jamming attacks

The classification of jamming attacks to GNSS takes into account several different parame-
ters, so, as we can see in Fig. 3.6, we can describe a jamming signal in the frequency and time
domain, with antenna directionality and also the specific jammer’s waveform modulation.
Given this parameters it is possible to define any possible jamming attack as we can see
from Table 3.1, where some examples of threats are listed taken from [28] and [29]:

• Spot jamming: occurs when a jammer focuses all of its power on a single frequency.

• Sweep jamming: the jammer attacks multiple frequencies one at a time in quick suc-
cession.

• Barrage jamming: also this attack threatens multiple frequencies, but, unlike sweep
jamming, does it simultaneously.

30
Figure 3.5: Positioning results around the true coordinates of U-Blox 5H receiver in a single-frequency jamming test.
[26]

Figure 3.6: Jammer parameters organized into trees. [27]

31
 Spot Sweep Barrage Systematic
jamming jamming jamming jamming
Full band × ×
Frequency Partial band
domain Single-tone ×
Multi-tone ×
Full time × × ×
Time
Partial time
domain
Impulses ×
CW × ×
Modulation Traditional data modulation
used by the AWGN ×
jammer Chirp × ×
Discrete frequency sweep
Table 3.1: Classification of specific jamming attacks.

• Systematic jamming: jamming attack synchronized with GNSS signals, with the in-
tention of causing maximum disruption with the minimum power expenditure.

3.1.2 Jamming detection

Clearly, given what I have described in the previous paragraph, what a GNSS user needs is a
means to detect the jamming attack, at least to give him/her a notification about the threat.
And, in the end, it all comes down to a binary hypothesis testing problem where it is necessary
to decide between

• H0 : absence of interference

y[n] = s[n] + w[n] for n = 0, 1, ..., N − 1, (3.1)

• H1 : presence of interference

y[n] = sIF [n] + vq[n] + w[n] for n = 0, 1, ..., N − 1, (3.2)

where w[n] is a realization of a zero-mean white discrete-time Gaussian noise W [n] with
variance σw2 , v is an amplitude factor, and q[n] is the IF digital version of the signal q(t) gen-
erated by a jammer.

32
Figure 3.7: Different approaches for jamming detection which can be implemented using measurements from differ-
ent receiver stages. [26]

In the hypothesis modelling here above, a decision is taken using N digital samples. A
general approach is to use such samples and construct a decision statistic D, designed on the
basis of, for one, a statistical model describing the behaviour of the digital samples in the
absence and in the presence of jamming. A decision between H0 and H1 is then taken by
comparing D with a decision threshold Th , which should be chosen such that a constant
false alarm rate is obtained. Since this model may be difficult to obtain, Th is often set using
criteria based on Monte Carlo simulations or on empirical results [26]. In Fig. 3.7 we can see
what is used to detect a jamming attack in different receiver stages.

3.2 Spoofing

By spoofing attack to the GNSS signals we consider the broadcast of false signals with the
intent that the victim receiver will misinterpret them as authentic signals. This may lead to
false position fixes, false clock offsets, or both, with the possible outcome that may be, for
example, that a hovering drone is sent into an unplanned dive by an attacker.
This type of attack, as one can expect, is more complicated since the spoofer must replicate
the RF carrier, PRN/spreading code and data-bits of each open-service GNSS signal that it
intends to spoof. Anyway, given the fact that the structure of most civilian GNSS signals is
known to the public and also since there has been advances over the years in software defined
radio (SDR) technologies, spoofing has become more feasible and less costly.
Spoofing signals can be generated with different techniques that can be more or less com-

33
 Authentic signals Shifted position
from satellites due to spoofing

GNSS spoofer

GNSS receiver GNSS receiver

(true position) (false position)

Figure 3.8: GNSS spoofing attack.

plex based on the result the attacker wants to achieve. A simple solution could be the use
of a GNSS signal simulator together with a RF transmitter which can send a forged signal,
that most likely won’t be synchronized with a legit one, to a GNSS receiver which will see
the new signal as noise. This, anyway, could be enough to negatively affect the acquisition,
or tracking, phase and degrade the receiver measurements. An example of a more complex
case is a receiver-based spoofer, which synchronizes itself with the current GNSS signals and
extracts the GNSS measurements. After this initial stage, the spoofer generates the spoofing
signal and sends it to the target’s receiver’s antenna.

3.2.1 Spoofing attacks

Based on the target, the spoofing attacks can be divided into signal processing and data level
attacks.

Signal level attack

As already said before, the structure of civilian GNSS signals, including the modulation
type, PRN signals, transmit frequency, signal bandwidth, Doppler range, signal strength
and many other features are publicly known. In this way a spoofer can generate forged sig-
nals similar to the one in Eq. (2.3) except for some spoofed parameters so as to effectively

34
 Lift-off Nonline of
Selective Jam and Trajectory
delay and Meaconing sight
delay Spoof spoofing
aligned spoofing
Amplitude × × × × ×
Delay × × ×
Doppler phase × ×
Table 3.2: Classification of specific spoofing attacks.

mislead its target receiver [30]

√
rs (t) = as 2Ps D(t − tp,s )C(t − τs ) cos[2π(fRF + fD,s )(t − tp,s ) + φ0,s ] + ns (t). (3.3)

In Table 3.2 we can see some possible signal processing level attacks from [31] and the descrip-
tion of what they modify.

• Lift-off delay: spoofer approaches the authentic signal with a relative delay ∆τI (t)
(and possibly Doppler), adjusting the spoofing signal’s power. It’s goal is to minimize
the delay and at the same time to increase its power so to move the tracking point
farther away from the true signal parameters.
• Lift-off aligned: as the lift-off delay, but it start with a ∆τI (t) near zero.
• Meaconing: it’s a fixed delay (∆τI (t) = constant) replica spoofing signal, that can
adjust its power levels.
• Selective Delay: these attacks are formed by estimating the code chip, replaying them
in delay to a user, and thus can be selective in terms of the satellites to be spoofed. In
case its power is smaller than the power of the real signal, it can be called multipath
attack as the spoofing signal looks like a nearby reflected signal.
• Jam and spoof : the spoofer forces the receivers into the acquisition mode using a jam-
ming attack that causes loss-of-lock on the authentic GNSS signals. Then the jammer
is switched off with the intention that the receiver acquires the spoofing signals.
• Nonline of sight spoofing: exploiting the fact that in suburban and urban environ-
ments a receiver will, in general, neither be able to track all satellites above the recom-
mended mask angle, nor will it be aware of the surrounding obstructions, the spoofer
can transmit signals only for potentially blocked satellites.
• Trajectory spoofing: a spoofer with the use of a software-defined radio or a GNSS signal
simulator attempts to capture the tracking points of all channels of a receiver along its
intended trajectory, forcing the user to follow the spoofed trajectory.

35
 Navigation data parameter Type of impact
Clock and ephemeris (CED) / time of Ranging errors (RE), position, velocity
ephemeris (ToE) data and time (PVT) errors
Health Denial of ranging
User / signal in space ranging accuracy Denial of PVT with receiver autonomous
(e.g. URA / SISA) integrity monitoring (RAIM)
Galileo system time (GST), coordinated
Timing errors
universal time (UTC)
GPS Galileo time offset (GGTO) Multi space vehicle RE
Data integrity, e.g. cyclic redundancy
Denial of navigation data
code (CRC)
Ionospheric corrections Ranging / PVT errors
Table 3.3: Highest Risk Navigation Data Parameters. [31]

Data level attack

Also the framing structure (e.g. almanacs and ephemeris) of a GNSS signal is publicly known
and, moreover, the information contained in it does not change in a short time (see Fig. 2.2
and Fig. 2.4). This allows the spoofer to take advantage of this period of time in order to
tamper the valid frame and have a different impact on the navigation message based on the
modified element (see Table 3.3). In 4.1.2 there will be the description of the software that was
created to modify a SDR simulator log file, which contains the single units of a navigation
message and is needed in order to create a navigation scenario.

It is clearly possible that both of the previous methods are implemented for a specific at-
tack, since the two types of attacks have different outcomes, and also jamming could be used
along with a spoofer to achieve better results.

3.2.2 Spoofing detection

To deal with the spoofing threat there exist several methods, but they can be divided into
two categories, predespreading and postdespreading [32]. In Tab. 3.4 we can see a summary
of the methods described in the following paragraphs.

36
Predespreading detection

The techniques that belongs to this detection method rely on the assumption that interfer-
ing signals are more powerful than the authentic ones. Predespreading methods evaluate the
overall power content of the received signal set without separately analyzing different PRN
signals. This category of detection looks for any abnormal variation in the received signal
power prior to the despreading process in the receiver. Some examples of this type of spoof-
ing detection metrics are

• Baseband variance analysis: it monitors the variance of baseband signals in order to

detect additional power injected by interfering signals. This method does not take
advantage of any spoofing signal features and simply assumes that the spoofing signals’
power content elevates the ambient noise floor.

• Structural power content analysis: it is a low complexity predespreading spoofing de-

tection approach that takes advantage of the cyclo-stationarity* of GNSS signals in
order to detect excessive amount of structured signal power in the received sample set.

Postdespreading detection

After the despreading process in the receiver, the used spoofing detection techniques rely
on the signal strength and its quality to identify the threat. A couple of examples of this
detection method are

• Effective C/N0 analysis: since the C/N0 monitoring is available in most commercial
receivers, this technique is a common signal strength monitoring metric. The cross-
correlation term caused by high power spoofing signals can become the dominant
term, so an abnormally high C/N0 value can be an indication of a spoofing attack.

• Signal quality monitoring, (SQM) the interaction between authentic and spoofing
signals causes distortion on the shape of the correlation function. SQM becomes an
excellent spoofing detection tool in the matched power spoofing scenario where all
PRNs are affected by spoofing.

* Signals that have statistical properties that vary cyclically with time.

37
 Anti- Receiver Spoofing
Spoofing
spoofing Complexity Effectiveness required scenario
feature
method capability generality
Baseband
Higher AGC
variance Low Medium Medium
power monitoring
analysis
Structural Specific pre-
power Higher despreading
Low Medium High
content power processing
analysis unit
Effective
C/N0
C/N0 Higher C/N0 Low Medium Medium
monitoring
analysis
Deviated
Signal shape of
Multiple
quality authentic Medium Medium Low
correlators
monitoring correlation
peak
Table 3.4: Summary of spoofing detection techniques. [33]

38
 GNSS threat analysis
4
In this chapter there is the description of the studies, the work and the testing that were
done during the internship carried out between August 2019 and February 2020.

4.1 Testbed for Attacks and Mitigations

In August 2019 I’ve started an internship to work on the Testbed for Attacks and Mitigations
(TAM) project, of which the Department of Information Engineering (DEI) is a subcontrac-
tor.
The goal of this project is to develop a tool to store, search and distribute services for vul-
nerability, and threat and mitigation identification for GNSS. The TAM will include both
technology and application specific threats, as well as real time location-based threats ob-
served in the territory for context awareness and emergency warning.
TAM is composed of the following key elements:

• IT infrastructure: a central physical infrastructure storing in a database all the vulner-

abilities and mitigations. It comprises all the information technology security needed
for the protection of information.

• Laboratories and control room and equipment: a number of laboratories and a control
room, connected to the infrastructure, which populate vulnerabilities and solutions.

39
 DC block
SDR
USRP Attenuation Receiver
Simulator

40dB

Clock

Figure 4.1: Equipment setup for the tests performed during the internship.

In addition, the TAM application will be used together with a specific GNSS simulator
to independently test the navigation/threat scenarios. For this reason, the ability to simulate
every possible threat to GNSS will be needed from this SDR simulator. So, besides the tests
needed to populate the database, also a software to emulate a data level spoofing attack, and
that had to be used together with this program, was proved necessary.

4.1.1 Equipment setup for the tests

To perform the various tests needed to fill the TAM database, we had to prepare the equip-
ment as can be seen in Fig. 4.1.

SDR Simulator The SDR simulator used for the tests is a real time GNSS multi-constellation
simulator, with advanced capabilities in terms of interference simulation. It is fully config-
urable for flexible generation of GNSS signals, interferences and authentication schemes up
to RF level. It runs in a laptop and, being fully software, enables a fast assessment of existing
and new emerging scenarios. It was used to create the navigation and threat scenarios to be
imported in the TAM database.

USRP Along with the simulator, the USRP was used to convert signals, to be sent to the
receiver, from digital to analog.

Clock The presence of an external reference oscillator improves the RF signal quality if
the internal clock is not stable enough.

DC block The DC blocks are passive coaxial components that prevent (block) the flow
of direct current (DC) frequencies to RF signals.

40
Figure 4.2: Power computed by the signal analyzer for a GPS signal from one satellite generated by the USRP.

Attenuation This component was needed to avoid receiver failure. To verify the atten-
uation needed by the commercial receiver, we used a signal analyzer to compute the USRP
output power (see Fig. 4.2). So, given that the used commercial receiver expected a gain of
20dB from the antenna, and since −67.35dBm are equal to −97.35dBW , the attenuation
a needed was

a(dBW ) = −97.35 − (−158.5) − 20 ≃ 40dBW,

where is −158.5dBW is the power of a GPS signal that reaches the Earth surface.

Receiver A commercial high precision multi-band GNSS receiver, that provides tracking
and measurements of all available GNSS signals (GPS, Galileo, GLONASS and BeiDou), has
been provided to carry out the tests.

4.1.2 Data level spoofing software

Given the lack of support for data level spoofing attacks in the SDR simulator, we had to
develop a software, to be used for the TAM project, to modify a log file of the simulator * to
emulate this type of attack. In this thesis, since the data structure for Galileo message is more
complex, but also mainly because the logic behind the classes needed to model the navigation

* A csv file which contains all the smaller units of a navigation message, in hexadecimal digits.

41
Figure 4.3: Columns added to the imported csv by EditDF class (random values in the fields).

messages† is similar, there will be described just the Python classes written for the E1b I/NAV
message.
First of all, some Python classes were written to address each component of a frame/sub-
frame/page in the Galileo navigation message. At first the csv is imported through the class
Csv2Message in a pandas DataFrame, a tabular data structure with labeled axes where each
of the functions to be used on the navigation message could be applied to groups of pages at
once quickly and easily. Then the class EditDF identifies which sub-frame/frame the pages,
that are ordered by TOW, belonged to, by simply using a counter and the modulo operation,
knowing that each sub-frame is composed of 15 pages and each frame by 24 sub-frames. As
last preliminary steps, some more columns are added to the original DataFrame table:

• "[bits]": contains the conversion from hexadecimal to binary using the Python
module bitarray.
• "bit0" ÷ "bit239": contain each bit of a page from "[bits]" column by num-
bering them from 0 to 239.

In Fig. 4.3 we can see the added columns by class EditDF to the original DataFrame, ob-
tained from the imported csv file, on a page composed of 60 hexadecimal random digits.
After this initial stage, there are several classes that deal with the modification of the dif-
ferent elements of a navigation message, and each one of them employs the BaseEdit class.
When a modification function is called, each class has its own way to obtain the needed pa-
rameters and then it calls a method of the BaseEdit class based on the type of edit wanted by
the user.
As regards the specific modification classes, after realizing that almanacs are elements that
are repeated every frame, ephemeris and clock corrections every sub-frame, and so forth,
the classes below were created with the following composition: a DataFrame containing the
name of the parameters that could be changed and the coordinates of those parameters in-
side of specific frames/sub-frames/pages, some getter methods to print a specific parameter
or a specific element (e.g. an entire almanac), and the different methods to edit the single
parameters.
†
The message composition was found in [34] for the GPS SPS, and in [20] for the Galileo I/NAV.

42
 • Almanac: used to modify the Almanac of one, or more, satellite.

• ArbitraryEdit: used to edit arbitrary parts of a page (without considering the specific
elements of each page).

• Corrections: used to manage time/ionospheric corrections and conversion parameters.

• Ephemeris: modifies the parameters of ephemeris.

• GstGpsConversion: for GST-GPS conversion parameters.

• OtherParams: for parameters that aren’t part of bigger entities and that are updated
in each subframe.

• OtherParamsInTwos: for parameters that aren’t part of bigger entities and that are
updated every two subframes.

• PageUpdate: is called every time the CRC has to be changed and is used to modify the
parameters of a single page.

Finally, a common element in each created class is the need for a method to update the
CRC field in the modified pages. So, the change_crc function was added at the end of
each edit method to allow the user to choose to update the CRC as it should be with the new
values, with one of the 3 BaseEdit methods, or by leaving unaffected that field.
Since each class of this software was designed almost with the same structure, except clearly
for the fact that different elements are updated with different rates, here below there is the
description of just the Almanac class as an example of all the classes that were written for this
task.

Almanac class

As regards Galileo I/NAV message, the almanacs are located into words 7, 8, 9 and 10, so, for
each subframe, they are located in the 4th and in the 5th pages. To receive the almanacs of
every satellite, an entire frame is needed, so every 2 subframes we have the transmission of 3
different almanacs. [20]
To describe this fact, a dictionary SV_ALMANACS was created by linking the parameter
name with its position in the navigation message. Then, to ensure ease of use, SV_ALMANACS
values were passed to a DataFrame.

43
 The first two methods of this class are two getters: one to get a single almanac parame-
ter, and the other for the entire almanac of a specific Space Vehicle (SV) in a specific frame.
Thanks to the DataFrame, only a couple of filters are needed to implement both gets.
The next step was to create a method, edit_support, to obtain a tuple of coordinates
to be used with edit methods. This function takes as input the common parameters of the
three edit methods, and returns a list containing the tuples with the coordinates for the edit.
This specific part of code is used for three similar functions that, for the sake of clarity, has
the same name of the three BaseEdit methods. All of them are written equally, except for the
BaseEdit method used by each one of them, so we have the computation of the coordinates
of the pages to edit followed by the editing, and, before the last row, the CRC update.
No editing method that was written has a return statement. This is due to the fact that
when a DataFrame is referenced, even if just a fragment of it, it is the same object we are
modifying and not a copy of it.

4.2 Threat modeling

In this section we describe how a threat is parameterized for insertion into the TAM database.
In particular, in order to describe jamming and spoofing signals, we created a model that takes
into consideration both the parameters needed by the SDR simulator software, and also the
parameters needed to model an attack as generic as possible. In the case of multiple jamming,
multiple spoofing, or jamming and spoofing together in the same scenario, each single attack
is modeled and then they are linked together in the simulation scenario.
Unlike the specific parameters that describe the configuration of an attack, those that de-
scribe the navigation scenario are the same for all attacks (see Table 4.1).

Simulation start time and duration Time and date (which are then converted re-
spectively in TOW and WN) of the simulation and how long it lasts.

Visible GNSS Which GNSS was/were used during the simulation.

Satellites used Which satellites were used for each GNSS which was selected.

Receiver Information about the model of the receiver, and the different receiver config-
uration settings used for the test.

44
Motion type If the vehicle where the receiver is installed is stationary or, otherwise, which
type of motion it should expect (e.g. terrestrial or airborne).

Parameter Value
Time and date February 04, 2020, 12:00:00
Duration 00:03:00
Visible GNSS GPS
Satellites used 1 ÷ 32
Receiver model receiver_a
PVT mode kalman
Multipath mitigation none
Motion type static

Table 4.1: Example of the key-value pairs needed to model a navigation scenario.

4.2.1 Jamming modeling

With regards to the jamming threat, the parameters described in the following paragraphs
were considered to effectively represent it (see Table 4.2).

Time of activation/deactivation Relative to the start of the test and indicates after
how many seconds the jammer starts/stops.

Jamming power The power of the jamming signal when the jammer is activated.

Jamming type The type of signal transmitted by the employed jammer (e.g. continuous
wave, Gaussian noise or chirp).

Jamming signal bandwidth The bandwidth of the jamming signal. If needed, it can
be also modeled a jamming attack that sweeps a certain frequency span with a given rate.

Impact description A brief explanation of the results from the test.

45
 Parameter Value
Time of activation [s] 90
Jamming power [dBW] -122
Jamming type chirp (sawtooth)
Impact description Small C/N0 loss

Table 4.2: Example of the key-value pairs needed to model a jamming attack.

4.2.2 Spoofing modeling

As regards the spoofing threat instead, the parameters needed to represent this type of attack
to the GNSS receivers were:

Time of activation/deactivation Relative to the start of the test and indicates after
how many seconds the spoofer starts/stops.

Gain offset The spoofing signal gain offset relative to the legit signal power.

Spoofing type The type of the implemented spoofing attack (e.g. fixed position or tra-
jectory spoofing).

Spoofed coordinates Single position, for the fixed position attack, or, otherwise, a set
of coordinates.

Impact description A brief explanation of the results from the test.

In Table 4.3 you can see an example of key-value pairs for the parameters needed to model
the spoofing attack.

46
 Parameter Value
Time of activation [s] 80
Gain offset [dB] 20
Spoofing type fixed position
Spoofed coordinates latitude and longitude values
Impact description Gradual shift towards the spoofed position

Table 4.3: Example of the key-value pairs needed to model a spoofing attack.

4.3 Tests and results

During the internship period, several tests were performed with the equipment setup de-
scribed in 4.1.1 to verify the reactions of the commercial receiver Open Service to different
types of threats. In the following sections there will be a list of parameters used for the tests
and a description of the outcomes for each of them.

4.3.1 Jamming tests

The goal for these tests was to verify, for each combination of GNSS used (GPS, Galileo or
GPS and Galileo):

• if some specific type of jamming was more dangerous than others for the receiver;

• the power interval [min, max] associated to a loss of accuracy of the attacked GNSS
signal. Below this interval no effect occurs, while, above this interval, the receiver loses
the track (i.e. the attack lead to a denial of service).

For the jamming tests, we set a test duration of 3 minutes and the attack is activated after 80
seconds. We have tried different jamming power, starting from the same power as the GNSS
signal and up to -105dBW, increasing from time to time by 20dBW, or until we experienced
a loss of accuracy. As jamming type, we tested

Continuous Wave Jamming The attack did not result in any problem for the receiver,
except for the maximum jamming power case where there was just a C/N0 loss (see Table 4.4).
We have assumed that these results were caused by some kind of implementation of a Notch
filter to mitigate this type of threat. In Table 4.5 you can see an example of parameterization
of this attack for a TAM database entry.

47
 GPS L1 C/A Galileo E1-B
min max min max
Jamming power needed to have loss of accuracy [dBW] - - - -
Table 4.4: Results from the CW jamming attacks.

Parameter Value
Time and date February 04, 2020, 12:00:00
Duration 00:03:00
Visible GNSS GPS
Satellites used 1 ÷ 32
Model receiver_a
PVT mode auto
Multipath mitigation none
Motion type static
Time of activation [s] 80
Jamming power [dBW] -105
Jamming type continuous wave
Impact description C/N0 loss on jamming activation
Table 4.5: Example of the key-value pairs needed to model a CW jamming attack.

Broad Band AWGN Jamming In this case we found out that at maximum jamming
power we have a denial of service, regardless of the GNSS. By decreasing the power (-111dBW
for the GPS or -112dBW for the Galileo) we could keep tracking but with a loss of accuracy,
a substantial C/N0 loss and a discontinuity in availability (see Table 4.6). In Table 4.7 you
can see an example of parameterization of this attack for a TAM database entry.

GPS L1 C/A Galileo E1-B

min max min max
Jamming power needed to have loss of accuracy [dBW] -117 -111 -118 -112
Table 4.6: Results from the broad band AWGN jamming attacks.

48
 Parameter Value
Time and date February 04, 2020, 12:00:00
Duration 00:03:00
Visible GNSS Galileo
Satellites used 1 ÷ 36
Model receiver_a
PVT mode auto
Multipath mitigation none
Motion type static
Time of activation [s] 80
Jamming power [dBW] -112
Jamming type AWGN
Availability discontinuity, loss of accuracy
Impact description
and substantial C/N0 loss

Table 4.7: Example of the key-value pairs needed to model a broad band AWGN jamming attack.

Narrow Band AWGN Jamming This attack was more successful than the broad band
one, since a smaller power was needed to cause a loss of accuracy and a denial of service. With
-122dBW of jamming power against a GPS signal we have a loss of accuracy, while, under -
119dBW, we had a definitive loss of lock with Galileo (see Table 4.8). Since there is a fast shift
between tracking the GNSS signals and having a denial of service, there was no substantial
C/N0 loss in these tests. As regards the parameterization, you can see the broad band AWGN
case since it is the same attack with a different jamming signal bandwidth.

Given that from the results of the CW jamming tests we hypothesized that the receiver
implements a Notch filter, we’ve also tested the AWGN jamming with increasingly narrow
band to see what is its upper bound. In the end we found out that the maximum bandwidth
where the noise is canceled is 500Hz.

49
 GPS L1 C/A Galileo E1-B
min max
Straight transition
Jamming power needed to have loss of accuracy [dBW] -122 -111 from tracking to
DoS
Table 4.8: Results from the narrow band AWGN jamming attacks.

Chirp (Sawtooth) Jamming As before, a smaller jamming power was needed to lose
tracking of Galileo signals (under -114dBW), while GPS still tracked, even if with a loss of
accuracy, with a jamming power of -110dBW (see Table 4.9). Under these values there was a
denial of service. In Table 4.10 you can see an example of parameterization of this attack for
a TAM database entry.

GPS L1 C/A Galileo E1-B

min max min max
Jamming power needed to have loss of accuracy [dBW] -121 -110 -117 -115
Table 4.9: Results from the chirp jamming attacks.

Parameter Value
Time and date February 04, 2020, 12:00:00
Duration 00:03:00
Visible GNSS GPS
Satellites used 1 ÷ 32
Model receiver_a
PVT mode auto
Multipath mitigation none
Motion type static
Time of activation [s] 80
Jamming power [dBW] -122
Jamming type chirp (sawtooth)
Impact description Small C/N0 loss

Table 4.10: Example of the key-value pairs needed to model a chirp jamming attack.

50
 Frequency Hopping Jamming As last jamming attack, we also tested frequency hop-
ping‡ . As for the CW jamming, for GPS we did not have denial of service, no matter the jam-
ming power. Anyway, we still had loss of accuracy on GNSS measurements up to -116dBW.
With regard to Galileo instead, we had loss of accuracy below -116dBW and loss of lock under
-111dBW (see Table 4.11). Above these values there was a denial of service. In Table 4.12 you
can see an example of parameterization of this attack for a TAM database entry.

GPS L1 C/A Galileo E1-B

min max min max
Jamming power needed to have loss of accuracy [dBW] -116 -105 -116 -112
Table 4.11: Results from the frequency hopping jamming attacks.

Parameter Value
Time and date February 04, 2020, 12:00:00
Duration 00:03:00
Visible GNSS GPS
Satellites used 1 ÷ 32
Model receiver_a
PVT mode auto
Multipath mitigation none
Motion type static
Time of activation [s] 80
Jamming power [dBW] -118.5
Jamming type frequency hopping
Impact description No effect

Table 4.12: Example of the key-value pairs needed to model a frequency hopping jamming attack.

4.3.2 Spoofing tests

Concerning these tests, they were made to see what would happen if we activate a fixed spoof-
ing attack without trying to cause a loss of lock with the legit GNSS signal. We tried with
‡
Method of transmitting radio signals by rapidly changing the carrier frequency among many distinct fre-
quencies occupying a large spectral band.

51
a gain offset of the spoofing signal, compared to the GNSS signal, from 0dB to 40dB in 3
different scenarios: a spoofed position at about 100m from the real position, then 1km and,
in the last one, 100km. As for the jamming tests, we set a test duration of 3 minutes and the
spoofing attack is activated after 80 seconds.

Fixed spoofed position at about 100m No matter the power difference with the
spoofing signal, there was no problem with the reception of the legit GPS signal. As regards
Galileo instead, even a 3dB gain offset was enough to gradually shift the position of the re-
ceiver toward the spoofed position (see Table 4.13). In Table 4.14 you can see an example of
parameterization of this attack for a TAM database entry.

GPS L1 C/A Galileo E1-B

Minimum gain offset needed for successful spoofing - 3dB

Table 4.13: Results from the fixed spoofing at about 100m attacks.

Parameter Value
Time and date February 04, 2020, 12:00:00
Duration 00:03:00
Visible GNSS GPS
Satellites used 1 ÷ 32
Model receiver_a
PVT mode auto
Multipath mitigation none
Motion type static
Time of activation [s] 80
Gain offset [dB] 20
Spoofing type fixed position
Spoofed coordinates position at about 100m from the receiver
Impact description No effect

Table 4.14: Example of the key-value pairs needed to model a fixed spoofing at about 100m attack.

Fixed spoofed position at about 1km In this test, the spoofed position is configured
1km away from the actual receiver position. Concerning the other parameters, the same con-

52
figuration considered for the previous test is exploited (see Table 4.14).
By increasing the spoofed distance we observed a loss in the C/N0 values for the GPS
signals from a 10dB gain offset. A 30dB gain offset was needed to effectively spoof the receiver
position. As regard the Galileo signal, the increased distance resulted in an higher gain offset
needed to spoof the legit position (see Table 4.15).

GPS L1 C/A Galileo E1-B

Minimum gain offset needed for successful spoofing 30dB 10dB

Table 4.15: Results from the fixed spoofing at about 1km attacks.

Fixed spoofed position at about 100km In this test, the spoofed position is config-
ured 100km away from the actual receiver position. Concerning the other parameters, the
same configuration considered for the 100m case is exploited (see Table 4.14).
In this case we had the same results of the 1km case for the GPS system, while, for Galileo,
there was again an increment to 30dB of the gain offset needed to spoof the legit position
(see Table 4.16). Probably this was caused by the fact that the receiver did not believe that it
moved so far. In the end, anyway, the gain offset was so overwhelming to cause a loss of lock
of the Galileo signal and to start the tracking on the spoofing signal.

GPS L1 C/A Galileo E1-B

Minimum gain offset needed for successful spoofing 30dB 30dB

Table 4.16: Results from the fixed spoofing at about 100km attacks.

4.3.3 Jam and Spoof tests

After the previous tests, we thought we would try a smarter attack, the jam and spoof (see
3.2.1). The attack was composed by:

• chirp (sawtooth) jamming: we verified with previous tests that this attack could pro-
voke a denial of service for both GPS and Galileo with a power of -105dBW;

• spoofing: we tried the same fixed position spoofing attacks that we tested before.

In particular, we increased the simulation time to 4 minutes and we organized that time
like this

53
 Time Description
00:00 Simulation start with GNSS signal transmission
01:30 Jamming attack
02:30 Spoofing attack
02:50 Jamming stop
04:00 Simulation stop
Table 4.17: Jam and spoof tests time organization.

From these tests we verified that if the spoofing attack succeeded with a specific gain offset,
then it would succeed also with higher gain offsets for sure.

Jam and spoof at about 100m As we expected, after 90 seconds, there was a denial of
service caused by the chirp jamming. After additional 60 seconds then, also the spoofer was
activated and, starting from a 0dB gain offset for the GPS, and from 3dB for Galileo, the
receiver started the tracking on the forged position (see Table 4.18). In Table 4.21 you can see
an example of parameterization of this attack for a TAM database entry.

GPS L1 C/A Galileo E1-B

Minimum gain offset needed for successful spoofing 0dB 3dB

Table 4.18: Results from the jam and spoof at about 100m attacks.

Jam and spoof at about 1km In this test, the spoofed position is configured 1km away
from the actual receiver position. Concerning the other parameters, the same configuration
considered for the 100m case is exploited (see Table 4.21).
We obtained that both GPS and Galileo just needed 3dB as spoofing signal gain offset from
the GNSS signal to start tracking on the spoofed position (see Table 4.20).

GPS L1 C/A Galileo E1-B

Minimum gain offset needed for successful spoofing 3dB 3dB

Table 4.19: Results from the jam and spoof at about 1km attacks.

Jam and spoof at about 100km This case needed an higher gain offset of the spoofing
signal power, compared to the legit one, to succeed. To spoof the real position for the GPS, a

54
10dB gain offset was enough. As regards the Galileo system, an higher gain offset was needed,
so we verified that with 30dB we could deceive the receiver (see Table 4.20). You can see the
parameterization for this attack in the 100m case (see Table 4.20).

GPS L1 C/A Galileo E1-B

Minimum gain offset needed for successful spoofing 10dB 30dB

Table 4.20: Results from the jam and spoof at about 100km attacks.

Parameter Value
Time and date February 04, 2020, 12:00:00
Duration 00:04:00
Visible GNSS GPS
Satellites used 1 ÷ 32
Model receiver_a
PVT mode auto
Multipath mitigation none
Motion type static
Jamming time of activation [s] 90
Jamming time of deactivation [s] 170
Jamming power [dBW] -105
Jamming type chirp (sawtooth)
Spoofing time of activation [s] 150
Gain offset [dB] 10
Spoofing type fixed position
Spoofed coordinates position at about 100m from the receiver
Spoofed position after the loss of
Impact description
lock caused by the jamming attack

Table 4.21: Example of the key-value pairs needed to model a jam and spoof at about 100m attack.

4.3.4 Jam and Spoof tests (shorter Jamming duration)

Since it is possible that a receiver may react differently, in a jam and spoof scenario, based on
the jamming attack duration, we deemed it appropriate to repeat the previous tests using a
different time of deactivation of the jammer.

55
 What we expected was that the receiver could possibly understand that it was not possible
for it to reach a certain distance in a shorter time. So, by decreasing the jamming attack dura-
tion, we should observe that a larger spoofing signal power gain offset is needed in order to
obtain a successful attack.
What we observed, from the tests we performed, confirmed our assumption. In particular,
we can see from Table 4.22 that, while on smaller distances (100m or 1km) the gain offset
needed was of 0dB or 3dB, on larger distances (100km) we need a much larger gain offset
to obtain the same results. This result may be caused by the fact that the receiver may not
believe that such distance may be covered in such a short time.

Jamming duration in the jam and spoof attacks

60s 30s 10s
GPS Gal GPS Gal GPS Gal
100m 0dB 3dB 3dB 3dB 3dB 3dB
1km 3dB 3dB 3dB 3dB 3dB 3dB
100km 10dB 30dB 20dB 40dB 20dB 40dB
Table 4.22: Results for the jam and spoof tests, on GPS L1 C/A and Galileo E1-B, with different jamming duration.

56
 Conclusion
5
In this thesis we showed the work that we did during the internship, in the context of the
TAM project. To better explain the topics we have dealt with during the internship, we also
presented a description of Global Navigation Satellite Systems, with a more focused look at
their signals and the threats that may harm them.
During the time spent at the company, we developed a software to simulate a data-level
spoofing attack. This software is able to address different civilian GNSS signals, specifically
the GPS SPS and the Galileo I/NAV. It can address every single parameter of a navigation
message and change it as required by the user. After the editing, it can then save the new
navigation message in a format convenient for spoofing simulations with the SDR simulator
software.
After that, some tests were performed to fill the TAM database with realistic threat sce-
narios. A parameterization of the navigation and threat scenarios was needed and we have
described the main parameters used to model specific attacks to the GNSS. Other parameters,
different from the ones in this work, are now present in the TAM database that are needed to
characterize a generic threat to the GNSS that may arise in the future. Anyway, they were not
shown because they were not required to effectively describe the tests that were performed.
From the tests results, we showed that the used commercial receiver was strong against
continuous wave jamming attacks, possibly for the presence of a Notch filter capable of fil-
tering an interfering signal with a bandwidth up to 500Hz. As regards other tested jamming

57
attacks, it turned out that narrow band jamming signals were more effective in disrupting
the tracking of the GNSS signals.
For spoofing attacks instead, the used receiver began to have problems even when we trans-
mitted the spoofing signal without a prior jamming phase. As expected then, since we already
had seen that the receiver was weak against this type of attacks, a smarter spoofing attack (jam
and spoof ) gave us the spoofed position as soon as the spoofer was activated.
For what concerns future work, it may be interesting to test the developed software to
change the navigation message fields with realistic scenarios and verify its impact on a receiver.
Besides that, clearly, more tests will have to be performed, with other types of attacks and
with different receivers, in order to make the database as exhaustive as possible. Moreover,
it may be interesting to find out which is the maximum distance, between the real and the
forged position, where the receiver may be deceived by our attacks, no matter the jamming
attack duration, using a small gain offset (0dB or 3dB).

58
 A
Error sources

GNSS measurements are affected by different types of errors that may be generated by
various sources: satellites, receivers and even from the signal propagation. [35]

A.1 Errors from satellites

Ephemeris errors The control segment predicts the position of satellites as a function
of time. Anyway, given that satellites can be affected by the asymmetry of the Earth’s gravi-
tational field, or the attractive force of celestial bodies, just to name a couple, these variations
may lead to differences of the position of the satellites with respect to the projected orbit.
This results in an information that is transmitted to the ground segment, that can generate
an error in the measurement of the satellite-receiver distance. Thanks to subsequent readings
done by the control segment, these errors can be reduced by using precise ephemeris.

Satellite clocks errors To provide frequency and time control requirements for
GNSS signals, atomic clocks are employed. Since these clock are not perfect, even if they
are highly precise, they produce errors in the order of meters, that, if we consider that a re-
ceiver’s clock is much less accurate, turns into an error in the order of hundreds of meters. By
acquiring the navigation signal from, at least, 4 satellites, the receiver may greatly reduce this
error.

59
A.2 Errors from receivers

Multipath errors Multipath is the main source of errors for phase and pseudorange
measurements, and it is due by receiving a signal reflected by, for example, a building. Since
the reflection causes a slight delay with respect to a signal that is received directly, the receiver
will compute an incorrect position. This error may be greatly reduced by placing the re-
ceiver antenna far from the reflective source. Anyway, current technology is able to provide
a stronger defense against this problem.

Receiver noise errors Receiver noise refers to the position error caused by the GNSS
receiver hardware and software. High-end GNSS receivers tend to have less receiver noise
than lower cost GNSS receivers.

A.3 Errors from signal propagation

Ionospheric delay The ionosphere contains electrically charged particles called ions
that delay the satellite signals and can cause a significant amount of satellite position error.
Ionospheric delay also varies based on the radio frequency of the signal passing through the
ionosphere, so GNSS receivers that can receive more than one GNSS signal (e.g. L1 and L2)
can use this to their advantage. By comparing the measurements for L1 to the measurements
for L2, the receiver can determine the amount of ionospheric delay and remove this error
from the calculated position.

Tropospheric delay Variations in tropospheric delay are caused by the changing hu-
midity, temperature and atmospheric pressure in the troposphere. Since tropospheric con-
ditions are very similar within a local area, the base station and rover receivers experience very
similar tropospheric delay. This allows Differential GNSS and RTK systems* to compensate
for tropospheric delay.

* Differential GNSS is a kind of GNSS Augmentation system, while Real-Time Kinematic systems use a
single base-station receiver and a number of mobile units to enhance the precision of the position data derived
from GNSS.

60
 References

[1] GSA GNSS Market Report, 2nd ed., European Global Navigation Satellite System
Agency, 2019.

[2] “Space Race,” https://en.wikipedia.org/wiki/Space_Race, accessed: 2019-11-22.

[3] K. Borre, D. Akos, N. Bertelsen, P. Rinder, and S. Jensen, A Software-Defined

GPS and Galileo Receiver, 1st ed., ser. Applied and Numerical Harmonic Analysis.
Birkhäuser Basel, 2017.

[4] J. Subirana, M. Hernandez-Pajares, and J. Zornoza, GNSS Data Processing, Vol. I:

Fundamentals and Algorithms, ser. ESA TM. ESA Communications, 2013.

[5] H. Kuusniemi, “User-Level Reliability and Quality Monitoring in Satellite-Based Per-

sonal Navigation,” Ph.D. dissertation, Tampere University of Technology, 06 2005.

[6] P. Teunissen and O. Montenbruck, Springer Handbook of Global Navigation Satellite

Systems, 1st ed., ser. Springer Handbooks. Springer International Publishing, 2017.

[7] “GSA Galileo,” https://www.gsa.europa.eu/european-gnss/galileo/

galileo-european-global-satellite-based-navigation-system, accessed: 2019-11-22.

[8] “Galileo Space Segment,” https://gssc.esa.int/navipedia/index.php/Galileo_Space_

Segment, accessed: 2019-11-22.

[9] “GPS Constellation status,” https://www.navcen.uscg.gov/?Do=

constellationStatus, accessed: 2019-11-22.

[10] E. Berger, “The last single-stick Delta rocket launched Thursday, and it put on a
show,” Ars Technica, 08 2019.

[11] “GPS Accuracy,” https://www.gps.gov/systems/gps/performance/accuracy/, ac-

cessed: 2019-11-22.

61
[12] S. Moore, “Superaccurate GPS Chips Coming to Smartphones in 2018,” IEEE Spec-
trum, 09 2017.

[13] I. Srivastava, “How Kargil spurred India to design own GPS,” The Times of India, 04
2014.

[14] “Galileo Constellation status,” https://www.gsc-europa.eu/system-service-status/

constellation-information, accessed: 2019-11-22.

[15] “BeiDou Constellation status,” http://www.csno-tarc.cn/system/constellation&ce=

english, accessed: 2019-11-22.

[16] G. Hein, “GNSS Interoperability: Achieving a Global System of Systems or ”Does

Everything have to be the same?”,” InsideGNSS, pp. 57–60, 02 2006.

[17] “Agreement on the Promotion, Provision and use of Galileo and GPS Satellite-based
Navigation Systems and related Applications,” 06 2004.

[18] “GPS International Cooperation,” https://www.gps.gov/policy/cooperation/, ac-

cessed: 2019-11-22.

[19] A. De Martino, Introduction to Modern EW Systems, 2nd ed. Artech House, 2018.

[20] European GNSS (Galileo) Open Service Signal-in-Space Interface Control Document,
1st ed., European Union, 12 2016.

[21] “An Introduction to GNSS,” https://www.novatel.com/an-introduction-to-gnss/

chapter-1-gnss-overview/section-1/, accessed: 2019-11-22.

[22] “GPS satellite trilateration,” https://giscommons.org/chapter-2-input/, accessed:

2019-11-22.

[23] “GPS Control Segment,” https://www.gps.gov/systems/gps/control/, accessed:

2019-11-22.

[24] L. Boccia, G. Amendola, S. Gao, and C.-C. Chen, “Quantitative evaluation of multi-
path rejection capabilities of gnss antennas,” GPS Solutions, vol. 18, 04 2014.

62
[25] S. Communications, “Spirent - Test Scenario Pack for GNSS Vulner-
abilities and Threats,” https://www.spirent.com/-/media/datasheets/
positioning/test-pack-gnss-vulnerabilities-and-threats.pdf?la=en&hash=
6B8125AF0B0E09B3778409CC35743DFED08FF889, 2018.

[26] D. Borio, F. Dovis, H. Kuusniemi, and L. Lo Presti, “Impact and Detection of

GNSS Jammers on Consumer Grade Satellite Navigation Receivers,” Proceedings of
the IEEE, vol. 104, no. 6, pp. 1233–1245, June 2016.

[27] M. Lichtman, J. D. Poston, S. Amuru, C. Shahriar, T. C. Clancy, R. M. Buehrer, and

J. H. Reed, “A communications jamming taxonomy,” IEEE Security Privacy, vol. 14,
no. 1, pp. 47–54, Jan 2016.

[28] A. Graham, Communications, Radar and Electronic Warfare. Wiley, 2011.

[29] J. T. Curran, M. Bavaro, P. Closas, and M. Navarro, “On the threat of Systematic
Jamming of gnss,” in Proc. ION GNSS, 2016.

[30] M. L. Psiaki and T. E. Humphreys, “Gnss spoofing and detection,” Proceedings of the
IEEE, vol. 104, no. 6, pp. 1258–1270, June 2016.

[31] R. T. Ioannides, T. Pany, and G. Gibbons, “Known vulnerabilities of global naviga-

tion satellite systems, status, and potential mitigation techniques,” Proceedings of the
IEEE, vol. 104, no. 6, pp. 1174–1194, June 2016.

[32] A. Broumandan, R. Siddakatte, and G. Lachapelle, “An approach to detect gnss

spoofing,” IEEE Aerospace and Electronic Systems Magazine, vol. 32, no. 8, pp. 64–
75, Aug 2017.

[33] L. Chiarello, “Security Evaluation of GNSS Signal Quality Monitoring Techniques

against Optimal Spoofing Attacks,” Master’s thesis, Università degli Studi di Padova,
2018.

[34] Global Positioning System Standard Positioning Service Signal Specification, 2nd ed.,
United States Government, 6 1995.

[35] “GNSS error sources,” https://www.novatel.com/an-introduction-to-gnss/

chapter-4-gnss-error-sources/error-sources/, accessed: 2019-11-22.

63
64
 Acknowledgments

I would like to thank prof. Nicola Laurenti, my supervisor for both my bachelor and mas-
ter studies, who got me into information security and inspired me with his vast knowledge
about this immense topic.
A special thank you goes to my closest friends Matteo and Martina. We shared the laughter
and the tears and I will forever be in your debt.
I am also grateful to all my friends that helped me relieve my stress during these years: the
ones with whom I roll dice every week, my colleagues from university, the ones that I have
known ever since I was little and the ones that I have met just a few years ago.

My deepest gratitude goes to my parents, Alessandro and Seila, and to my girlfriend, Dom-
iziana. You kept me going on and this work would not have been possible without you.

65