Computer Science Question

below r the document

Received February 8, 2021, accepted March 9, 2021, date of publication March 15, 2021, date of current version March 23, 2021.
Digital Object Identifier 10.1109/ACCESS.2021.3065926
Survey on Anti-Drone Systems: Components,
Designs, and Challenges
SEONGJOON PARK , (Graduate Student Member, IEEE), HYEONG TAE KIM , SANGMIN LEE,
HYEONTAE JOO , AND HWANGNAM KIM , (Member, IEEE)
Department of Electrical Engineering, Korea University, Seoul 02841, South Korea
Corresponding author: Hwangnam Kim (hnkim@korea.ac.kr)
This work was supported by the National Research Foundation of Korea funded by the Korean Government under
Grant 2020R1A2C1012389.
ABSTRACT This paper presents a comprehensive survey on anti-drone systems. After drones were released
for non-military usages, drone incidents in the unarmed population are gradually increasing. However,
it is unaffordable to construct a military grade anti-drone system for every private or public facility due
to installation and operation costs, and regulatory restrictions. We focus on analyzing anti-drone system
that does not use military weapons, investigating a wide range of anti-drone technologies, and deriving
proper system models for reliable drone defense. We categorized anti-drone technologies into detection,
identification, and neutralization, and reviewed numerous studies on each. Then, we propose a hypothetical
anti-drone system that presents the guidelines for adaptable and effective drone defense operations. Further,
we discuss drone-side safety and security schemes that could nullify current anti-drone methods, and propose
future solutions to resolve these challenges.
INDEX TERMS Anti-drone, counter-drone, drone detection, drone identification, drone neutralization.
I. INTRODUCTION
Advances in micro air vehicles, also known as drones, take
advantage of opportunities in the several industrial domains,
from agricultural engineering to military missions [1]. Rapid
expansion of the drone industry has surpassed regulations
for safe and secure drone operation, which makes them representative means of the illegal and destructive terrors and
the crimes [2]. With the introduction of drones into civilian
technology, drones are now gaining attention as a threat
of safety and security, which leverages the emergence of
the anti-drone (or counter drone) technologies. Anti-drone
systems are devised to defend against drone accidents or
terrorism, and needed to be advanced to cope with the future
drone flight systems.
Currently, most of anti-drone systems adopt military grade
components to achieve the confirmatory destruction of malicious drones. However, several difficulties apply when locating military grade anti-drone system into civilian areas.
Military counter-drone measures typically use jamming systems [3] to disable the target drone control channel. The
The associate editor coordinating the review of this manuscript and
approving it for publication was Sara Pizzi
VOLUME 9, 2021
.
jammer generates extremely high amplitude of RF signal
in the target frequency band to prevent communication. For
military scenarios, the site is controlled by the military, and
the operator pre-acquires proper field manual for jamming
conditions, hence the side effect of jamming can be ignored
or managed. However, for non-military applications, RF jamming to neutralize high-speed drones risks temporal paralyze
of existing wireless network systems, such as mobile access
or wireless sensor networks. Thus, most national regulations
prevent non-military use of jamming systems [4], [5], and
hence civilian anti-drone systems need to investigate alternative approaches to stop illegal or unauthorized drones.
Similarly, anti-aircraft weapons such as missiles are hardly
allowed for civilian systems. Except for cases where the
national army covers the entire civilian area, such as Iron
Dome [6] in Israel, non-military operators need anti-drone
strategy without using military grade weapons.
Radar was regarded as a limited solution for drone detection due to inflexible radar cross sections (RCS) [7], but
recent radar technology advances enable arbitrary drone
detection with acceptable identification rate [8]. Thus, radar
is becoming adopted for long range drone detection [9], but
its use also suffers from national regulations, such as RF
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
42635
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
license policies [10]. The difficulty and relatively high cost
of installing drone detection radars makes civilian counter
drone systems look for other drone detection methods, such
as vision [11] and RF signal [12] systems.
Civilian drone stopping strategies tend to employ unarmed
methods such as hijacking (Section V-A) or capturing
(Section V-F) solutions. These methods are technical counterpoint of drone’s safety and stability systems, and both sides of
methodologies are equally on demand in drone research. As
a breakthrough in such a competitive situation, it is essential
to refine the anti-drone system in a structural manner in order
to cope with the drone’s defense mechanism by adaptively
responding to the drone’s avoidance strategy. To do so, stateof-the-art drone security and safety studies should be evaluated, not only attempt to take advantage of the conventional
drone mechanism.
This paper studies non-military anti-drone systems in
comprehensive way. Considering recent drone incidents,
we specially investigate the requirements for non-military
anti-drone systems. We do not only list suitable methodologies, but propose guidelines for anti-drone system design that
efficiently merge the components. Finally, we provide milestones to advance counter drone technology against drone
security evolution.
A. ROADMAP FOR THIS PAPER
Fig. 1 graphically represents the organization of this paper,
with the following details.
Section II discusses anti-drone system motivations and
objectives for drone attack cases over the last few decades.
Section II-A lists recent non-military drone incidents, highlighting safety and security awareness of malicious drones.
Section II-B identifies requirements and breakthrough for
applicable anti-drone systems, developing the major criteria
to evaluate present system components and design anti-drone
system guidelines.
Sections III–V introduce anti-drone components, divided
into detection, identification, and neutralization phases,
respectively.
Then, Section VI considers actual anti-drone system
installations to survey current usage and further system extension requirements. Section VII proposes guidelines anti-drone system design, installation and operation.
Sections VII-A–VII-C address detection system deployment,
methods to evaluate drone attack situations, and where to
neutralize illegal drones.
Section VIII considers future aspects and aims for
anti-drone systems. Section VIII-A introduces drone-side
safety and security methods against anti-drone technology,
which can nullify attempts to detect and neutralize illegal drones. Section VIII-B consequently derives anti-drone
development directions to sustain robust defense against
malicious drones.
To avoid semantic confusion, each section defines
anti-drone terminologies commonly used in the domain. Frequently used in anti-drone terms sometimes have ambiguous
42636
scope, e.g. hijacking, spoofing, and jamming. Hijacking and
spoofing are often used interchangeably, and jamming sometimes includes drone neutralization methodologies, or only
drone communication interruption solutions. Clearly defining
these terms helps to avoid contextual conflicts and fit with
anti-drone system operator requirements.
To the best of our knowledge, this study is the first one
investigating non-military grade anti-drone systems. This
paper provides a useful survey for non-military anti-drone
system and contributes to future technology developments.
II. ANTI-DRONE BACKGROUNDS
This section considers the motivations and requirements of
anti-drone system. Drone industry expansion has increased
injudicious, unauthorized, and illegal drone use, causing considerable social and economic damage. We review some
major drone incidents worldwide, and derive essential features for emerging anti-drone systems.
A. DRONE INCIDENTS
Drone illegal use and terrorism have recently occurred in
various ways. We list and analyze several key incidents to
derive appropriate anti-drone system objectives.
1) ILLEGAL FLIGHTS AT AIRPORTS
Gatwick Airport, the second largest airport in the UK, was
paralyzed for a day in December 2018, by an illegal drone
that breached the runway airspace [13]. Illegal drones have
appeared near the airport more than 50 times for almost
15 hours. This happening seemed to be intentional to confuse
the airport operations, since these are industrial drones and
considerably larger than commercial models. Illegal drones
also appeared near Frankfurt Airport, Germany, in May 2019,
closely at the aircraft landing area for approximately one
hour [14]. In both cases, the drones reached at important
locations such as runway or crucial airspace without being
detected by any of airport security systems. These incidents
occurred large ecnomic loss due to poor detection distance
and accuracy, and lack of response1 procedures against unauthorized or illegal drones.
2) ATTACKS ON PUBLIC INSTITUTIONS
An unmanned aircraft equipped with a C-4 bomb attempted
to attack the U.S. Department of Defense and Capitol Hill in
September 2011 [15]. Fortunately, the criminal was arrested
by the FBI before explosion. This was the first known terrorism using drones and an example of FBI terrorist prevention
through tracking and proactive blocking. The case highlights
that building an anti-drone system is important in practical
terms, but requires cooperation with national organizations
such as police and military.
The Russian military defended the first drone fleet attack
in January 2018 [16]. Thirteen armed fixed-wing unmanned
1 We use a term response to meaning combined detection, tracking and
neutralization.
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
FIGURE 1. Roadmap for this paper.
aircraft deployed to attack Khmeimim Air Force base and Tartus naval installation, but repelled by Russian military radio
electronic warfare technology. Ten drones were shot down
by missiles, and the other three were blocked by Russian
hijacking technology. Several military bases had high level
anti-drone systems, but anti-aircraft systems such as missiles
cannot be used in non-military site. Thus, anti-drone systems
should prepare neutralization technologies without weapons,
such as hijacking and capturing.
At Aramco, Saudi Arabia’s national oil company, the
largest oil refining facilities were burned and shut down by
drone attack in September 2019 [17]. Ten drones attacked
the facility, carrying 3 kilogram of explosives per unit. The
incident caused huge damage to Saudi Arabia’s crude oil production and peak price of international crude oil. This attack
succeeded due to lack of simultaneous detection and defense
systems for multiple drones. However, it is almost impossible
to install drone neutralization equipment to completely cover
such large number of facilities and enterprises. Therefore, it is
essential to prioritize and concentrate anti-drone systems in
key facilities.
3) ATTACKS ON SPECIFIC INDIVIDUALS
A small drone containing radioactive materials was dropped
on the roof of the Japanese Prime Minister’s residence in
April 2015 [18]. Not only was the drone able to fly to
the Prime Minister’s residence, it was left unattended for
approximately two weeks. Clearly, there was poor or no
drone detection system installed. However, it might have been
VOLUME 9, 2021
difficult to install intensive detection equipment due to the
circumstances of the location, particularly privacy. Therefore,
it is essential to secure various detection methods to fit to the
area’s requirements.
The Islamic militant group Islamic State (IS) has been
using small drones to drop grenades since 2016. IS killed two
Iranians in Syria in October 2016 with two ultra-small drones
purchased from Amazon [19]. This was considered to be the
first case of terrorism using commercial drones, and the case
is important in that IS used commercial off-the-shelf drones,
establishing that a wide range of drone terrorism was possible
because the drones could be easily obtained without requiring
expert-level skill to fly them.
Two drones equipped with bombs attempted to assassinate Venezuelan President Nicolas Maduro at a national
outdoor event in August 2018, but failed [20]. This was
the first attempt to use a drone to assassinate the head
of the country. This case of incident highlights the need
for anti-drone systems in the cases of temporal events. To
cope with these portable scenarios, temporary anti-drone
systems require rapid installation and deployment of their
equipment.
Various drone incidents using small drones are difficult to
detect, regardless of the type of sites, military or non-military.
Illegal drones are mainly to paralyze major facilities [13],
[14], [17], terrorist attacks [15], [16], or attacks on specific
people [18]–[20]. In addition to the listed cases, there are
numerous cases of minor accidents such as restricted area
invation by unauthorized or illegal drones. The demands on
42637
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
the anti-drone system to prevent such incidents are exploding
worldwide.
B. ANTI-DRONE SYSTEM REQUIREMENTS
From the observations in Section II-A, we summarize core
requirements for anti-drone systems as follows.
Drone-specialized detection. Conventional zone security systems include drone-detection equipment such
as radars or cameras, but lack the performance and
awareness to allow current systems to recognize various
drone incidents. When designing anti-drone system with
current monitoring equipment, the overall architecture
should be revised to detect various drones at sufficient
distance to prepare the defense.
• Multi-drone defensibility. Some previous illegal drone
incidents [16], [17] highlight the potential for a drone
fleet attack. Sooner or later, various numbers of (legitimate) flying objects including personal air vehicles (PAVs) would be around the area, which leads to
situations where multiple drone threats will need to be
simultaneously detected and handled.
• Cooperation with security organizations. Seizing and
intercepting drone threats such as [15] is the prime way
to safely defend an area against unauthorized or illegal
drones. In addition to the preemptive investigation, regulatory restrictions and cooperation opportunities with
national or public security systems (such as police or
military) should be discussed.
• System portability. As shown in II-A3, defending an
area against unauthorized or illegal drones can vary
depending on space and time. Immediate anti-drone
deployment can be accomplished with mobilized detection, identification, and neutralization components,
which require lightweight equipment and competent
wireless networks.
• Non-military neutralization. There has only a single
successful defense against drone attack reported [17],
which was possible by deploying military grade
weapons. Although drone jamming has been largely
adopted and tested for commercial anti-drone systems,
jammers could not stop the physical threat of the uncontrolled drones. Thus, a definitive neutralization methodologies and procedures are required.
•
Fig. 1 shows a typical anti-drone system comprising multiple subsystems. Anti-drone research domain remains in
early-stage development, in contrast with drone stabilization technologies [21]. Solid solutions such as jamming or
anti-aircraft weapons provide acceptable results for demands
on stopping the drones, but place a heavy burden on regulations and financial budget. Therefore, we focused on
approaches that could be deployed for non-military grade
facilities, such as civil airports, sports stadia, outdoor/indoor
convention sites, etc. Sections III–V list anti-drone system
component surveys, and evaluate each considering the above
requirements to build an effective anti-drone system. To have
42638
global view of the domain from research to product, we comprehensively surveyed vendor catalogues, articles, and white
papers as well as research papers.
III. ANTI-DRONE SYSTEM: DRONE DETECTION
Drone detection exploits various features of flying drone.
Drones commonly emit heat, sound, and RF signals to communicate with the remote operator. Detection system collects
sensor data to confirm the presence of drones in nearby
areas. Depending on the measure, it can speficy the drones’
expected locations.
Table 1 shows drone detection schemes categorized by
sensing technology. The following subsections consider each
detection strategy and explore the basic mechanism and technical limitations.
A. THERMAL DETECTION
Physical components such as motors, batteries, and internal hardware radiate significant amount of heat, which can
be recognized by thermal cameras [26]. Many studies have
proposed detecting target drones by their heat signatures.
Andraši et al. [22] proposed a drone detection scheme to
detect thermal energy emitted by the drone during flight.
Wang et al. [56] employed a convolutional neural network to
enhance the system performance and accurately detect target
drones from thermal images. The Spynel [23] product from
HGH Infrared Systems detects infrared from the object heat,
enabling 360◦ surveillance.
Thermal detection has advantages in terms of weather
resilience, identification availability, and lower cost than
radar based systems. However, the practical detection
range (51 m [22]) is considerably shorter than most other
approaches, hence enhancing granularity of detection scheme
or improving resolution of thermal imaging camera are major
challenges.
B. RF SCANNER
Drones controlled by an operator usually exchange specific
messages as RF signal containing sensor output, flight commands, etc. RF scanner technologies capture wireless signals and determine the existence of drones in the target
area. Signal intelligence (SIGINT) and communication intelligence (COMINT) are primitive models for RF based drone
detection. Al-Sa’d et al. [28] designed a drone RF signal
learning and detection system using a deep neural network
with multiple hidden layers to categorize detected drone types
and flight modes. Although classification accuracy decreases
with increasing number of classes (drone types), detection
accuracy is acceptable. Da-Jing Innovations (DJI) released
Aeroscope [57], a detection system that collects DJI drone
control data at around.
As discussed above, the major disadvantage for RF based
detection is that it cannot detect drones that do not exchange
RF signal continuously, such as the ones in autonomous
navigation. In addition, since RF scanner detects the drones
by signal analysis, drones using unknown control protocols
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
TABLE 1. Drone detection technologies.
FIGURE 2. Frequency modulated continuous wave (FMCW) mechanism.
or different frequency bands [12] are challenging to detect.
Nevertheless, most drone detection systems use RF scanner,
due to its long range and low cost, while combining with other
methodologies.
1) RADAR BASED DETECTION
Radar detects physical objects and determine its shape, distance, speed, and direction by sensing reflected Radio signals.
In contrast with RF scanner, radar measures time-of-flight for
the reflected signal, whereas RF scanner demodulates the signal itself. Continuous-wave radar characteristically measures
target velocity using range and Doppler information.
Fig. 2 shows a typical radar based detection system.
Frequency modulated continuous wave (FMCW) radar and
coherent pulsed Doppler radar retain and track transmitted
and received signal phases to estimate distance and velocity.
In Fig. 3, FMCW radar derives distance R from speed of light
c; and multiple measurements of δt; Doppler frequency shift
fD ; and bandwidth BW . Then, the velocity of object can be
calculated from c, wavelength λ, and angular deviation θ [58].
Radar surveillance and tracking uses several frequency
bands [59], [60], which we summarize below.2
• Ka, K, and Ku bands, above 18 GHz, very short
wavelength. Used for early airborne radar systems,
but uncommon today except maritime navigation radar
systems.
2 This paper refers to nominal frequency range for each band, and specific
frequency ranges differ from international/national workgroups.
VOLUME 9, 2021
FIGURE 3. Distance and velocity determination by FMCW.
X-band, 8–12 GHz. Used extensively for airborne systems for military reconnaissance and synthetic aperture
radar.
• C-band, 4–8 GHz. Common in many airborne research
systems (e.g. CCRS Convair-580 and NASA AirSAR [61]) and spaceborne systems (e.g. ERS-1 and
2 and RADARSAT [62]).
• S-band, 2–4 GHz. Used for Russian ALMAZ satellites
and weather radar.
• L-band, 1–2 GHz. Used for US SEASAT and Japanese
JERS-1 satellites and NASA airborne systems.
• P-band, 300 kHz to 1 GHz. Longest radar wavelengths,
used for NASA experimental airborne research systems.
•
Radar is also classified into 2D and 3D by the type of the
phase array antenna [63]. 2D radars adopt passive electronically scanned array antennas (PESAs), which control beam
steering by electric field phase applied to each array element,
providing relatively large detection range while wideband
utilization is not possible. 3D radar commonly uses active
electronically scanned array antennas (AESAs), which control beam steering and shape by the electric field gain and
phase of each element. Although AESAs have relatively short
detection range, they can self-correct errors and support wideband detection. Several studies have implemented 3D radars,
e.g. [64]. The main difference between 2D and 3D radar
is that 3D radar can estimate the altitude of target objects,
42639
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
whereas 2D radar acquires limited information of z-axis
through auxiliary systems [65], [66]. 3D radar is desirable for
anti-drone systems, but 2D radar with other methods can be
a better solution from the view of large-scale monitoring and
cost efficiency.
Although radar has been widely adopted for military and
civil surveillance systems [67], early drone detection systems
were skeptical about using radar, due to extremely low drone
RCS [68]. Liu et al. [34] proposed multi-channel passive
bistatic radar (PBR) to improve radar detection granularity, correcting the drone’s location by extended Kalman filter (EKF) and global nearest neighbor (GNN) approaches.
Several drone detection studies proposed high resolution
FMCW radar with various improvements, including phase
interferometry, functional modes, and various bands [35],
[69], [70].
Radar based drone detection offer longer detection range
and constant observability compared with RF scanner, but
there are some detection availability and regulatory limitations. Radar cannot distinguish a drone from obstacles if the
drone hovers in one position or flies at low speed. Thus,
combining radar and other technologies (camera, RF scanner,
etc.) is strongly recommended. Radar systems also continuously emit high power RF signals, so nation permission
is required for frequency bands and installation locations.
In particular, facilities that already operate radars, such as
airports [71] may have difficulty installing additional radars
due to RF interference issues. Partial spectral overlap between
radar and radio waves can cause bad signal interference and
poor performance of both radar and the network. Several
studies investigated mutual interference between radar from
military or other government/private organizations and radio
access networks such as 5G to ensure coexistence [72]–[75].
Administrator should consider these RF circumstances in
anti-drone system installation.
C. OPTICAL CAMERA DETECTION
Similar to thermal camera detection, optical cameras for
drone detection have been widely investigated for anti-drone
application. Sapkota et al. [42] exploited histogram of oriented gradients features to detect drones from captured
images, and Jung et al. [76] proposed a video based drone
surveillance system to monitor large 3D spaces in real time.
Drone detection equipment based on optical cameras provide
extremely low cost and less regulatory limitations than previously discussed ones, enabling fine-grained tracking system
via dense deployment. However, the shortcomings including
relatively short ranges, high weather dependency, and impermeability to obstacles force the fusion with different sensing
systems. Widely adopted military electro-optical/intra-red
(EO/IR) systems combine optical cameras and infrared sensors for drone detection [77].
D. ACOUSTIC SIGNAL DETECTION
Drone detection sensing acoustic signal emitted from the
motors [78] directly exploits an inherent drone feature.
42640
Kim et al. [47] proposed plotted image machine learning
and k-nearest neighbors, achieving 83% and 61% accuracy,
respectively. Aside from short detection range, direction measurement and drone tracking are remaining challenges.
Fig. 4 compares drone detection components with respect
to their functionalities and detection ranges. As shown in the
figure, radar achieves high amount of minimum detection
range, due to its inherent mechanism [79]. Most vendors
propose hybrid drone detection systems for availability, accuracy, and installation flexibility. Some vendors provide automatic systems combining both detection and neutralization,
commonly targeting and jamming, but the jammer use is
highly limited in most countries. Thus, non-military systems
need to fine-tune a wide range of requirements including
jamming limitations, relevant existing radar installations, and
drone neutralizing techniques considered in more detail in
Section VII.
Table 2 presents the availability of drone detection technologies for problems that may interfere. The table and Fig. 4
explicitly show that each method cannot perfectly satisfies
current requirement of anti-drone system. To break this limitation, drone detection system should be designed in a cooperative way that combines the clues from multiple equipment.
To do this, not only each method should be improved to
enlarge the cover area in Fig. 4, multiple mechanism should
be combined as a hybrid system, considering the security
requirement of defending area. For instance, RF scanner
has big advantages in both range and functionality, except
for the limitation that can only be used for commercial
drones. Thus, RF scanner is acceptable in large scale area
for detecting hobby drones flying in illegal. Meanwhile,
the drone-sensitive spots where precise tracking of any flying
objects is essential, such as airstrips or nuclear piles, must be
equipped with detection components including vision, radar,
and acoustic. To cope with non RF-detectable drones such
as terrorist drones, high security area should locate versatile
detection methods for preventing drone concealment technologies (Section VIII-A). In Section VII, we address some
guidelines to deploy drone detection system with examples.
E. HYBRID DETECTION SYSTEM
Sections III-A–III-D show that using a single detection
method inevitably results in drone detection blind spot, which
makes it difficult to successfully neutralize illegal drones.
Most vendors install hybrid drone detection systems that
employ sensor fusion technology and joint hardware control.
We discuss some cases of their hybrid schemes.
•
Radar + vision. Radar and optical (or thermal) cameras
provide excellent complementarity for drone detection.
Vision based detection can easily track drones by controlling image zoom, tilt, and focus, but struggles with
dynamic control over the target area; whereas radar
detection provides omnidirectional wide area scanning
with low drone identification and low scan frequency.
Thus, radar scans the target area, and vision system
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
FIGURE 4. Detection method classification with respect to functionality and range.
TABLE 2. Detection technology availability.
controls external and internal camera parameters to
accurately investigate suspicious points. This combination dynamically compensates for each other’s flaws,
and hence many vendors adopt this structure [80]–[83].
• Multiple RF scanners. RF scanners can detect drones
and additional information (type, control commands,
and so on), but not always their location. If the drones
are controlled only by pulse position modulation (PPM)
or pulse width modulation (PWM) messages, they may
not emit location information on an RF channel. Fig. 5
shows multiple RF scanners receiving RF message and
calculating drone locations by traditional RF localization schemes [49]. Since RF scanners are generally
cheaper than equivalent coverage of radar systems, some
vendors dominantly use multiple RF scanners instead of
radar [84].
• Vision + acoustic. Combining vision and acoustic sensors is a traditional sensor fusion technique to improve
detection accuracy [49], [85]. Vision based detection
struggles to distinguish unfamiliar drone shapes, and
acoustic based detection achieves low performance
in noisy environments. The complementary design is
effective in terms of weather resilience, environmental
VOLUME 9, 2021
FIGURE 5. Drone position tracking by multiple RF scanners.
resilience, and detection accuracy, hence [81], [83] products commonly employ this design.
Table 3 lists several commercial anti-drone system components. Various detection technology combinations are available to achieve similar 3–5 km detection range. Multiple
systems can be installed, e.g. [82], with reliable and
42641
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
TABLE 3. Hybrid detection systems.
low-latency networks. Thus, finding an efficient detection
configuration for the target area is an essential step for constructing a robust anti-drone system.
IV. DRONE IDENTIFICATION
Before listing identification systems, we clarify detection and
identification terms used in this paper to avoid confusion.
Drone detection refers to systems that observe a flying (or stationary) object and determine if the object is a drone, whereas
drone identification refers to determining if the detected
drone is illegal and hence should be neutralized. For example, a minimal radar system may barely accomplish drone
detection, since it cannot distinguish between drones and
similar sized birds without an additional prediction scheme
or auxiliary equipment (e.g. vision cameras) as discussed in
Section III-E. Identification should be performed accurately,
robustly, and promptly; particularly where the target area
utilizes drones or permits legal use of leisure drones. The
identification system should cooperate with the detection
system to defend the target area without false neutralization.
Ideally, drone identification should passively identify the
legality of drones through identification tags attached to them
that periodically broadcasts their information, such as RFID
tags. However, comprehensive attachment of tag can be discussed nationally or internationally, and has currently just
begun. Furthermore, unintended or inexperienced control of
legal drones can also be a threat to nearby facilities. Thus,
any anti-drone system should include active identification
solutions to determinate hazard level for detected drones,
by tracking and estimating flight paths, and collecting specific information such as drone model and detailed properties
that violate safety regulations. As described in Section 4,
some detection systems can also provide identification
functionalities, such as DJI aeroscope [57], or fine-grained
detection network which can track the flight path. This
section discusses active and passive drone tracking and path
estimation solutions, mainly based on vision techniques.
A. DRONE TRACKING AND FLIGHT ESTIMATION
To distinguish detection methods, we focus on theoretical
and systematic location estimation schemes for flying drones.
Most systems use vision information to improve drone
42642
tracking, using conventional image processing or machine
learning (convolutional neural networks). Path estimation
systems also use neural networks or various filters over
the tracking results to determine the drone movements.
Xie et al. [87] improved the particle filter algorithm to more
precisely estimate drone location from measured azimuth,
elevation, and distance between the drone and the detection equipment, and modelled drone constant acceleration.
Son et al. [88] proposed an optical flow based tracking
method to track fast and small drones. The authors combined
a recursive filter to detect tracking failures caused by fast
positional changes and perform retracing. Xue et al. [89] proposed a multi-layer neural network for drone path estimation
which approximates any continuous function in a specified
space to estimate dynamic non-linear drone movement, and
determined the network parameters. Drone tracking and position (or motion) estimation may be insufficient to judge drone
legality, but it is essential to estimate how much drone motion
could threaten the defense area. We introduce path-based
drone threat assessment in Section VII-B2.
B. RFID BASED IDENTIFICATION
Radio frequency identification (RFID) has been widely
adopted for identification and real-time location systems (RTLS) in recent decades [90]. Active RFID system is
a promising drone identification approach due to low cost
and lightweight system design. Buffi et al. [91] proposed
an RFID based drone identification system over large areas,
to distinguish between licensed and unlicensed drones. The
major challenge is range extension and security concerns.
High speed drones may not give a short range RFID system
sufficient time to identify them. Spoofing RFID signal can
also deceive the system and allow malicious drones to trespass over the defended area. Thus, security schemes for RFID
drone identification systems [91] and long-range active RFID
communication [92] should be further studied.
In addition to identification, positioning RFID-tagged
drones has also been widely studied for drone tracking.
Choi et al. [93] proposed a differentiated method for indoor
localization by attaching Ultra High Frequency (UHF) RFID
tags to UAVs and installing a number of passive tags over the
target area, connected to the system. This solution exploits
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
interference between UAV and ground tags, which can be
detected by measuring the received signal strength indicator
(RSSI). The system first measures the tag’s RSSI variance,
and then dynamically estimates drone future location by finding the spot with greatest signal interference comparing with
pre-measured variances. Locating wired passive tags over
large areas may be cost-intensive, but position estimation in a
coarse distribution of the ground tags can extend the available
area.
C. AUTOMATIC DEPENDENT SURVEILLANCE BROADCAST FOR DRONES
Automatic dependent surveillance – broadcast (ADS-B) has
been adopted for aircraft Air Traffic Control (ATC) systems [94]. ADS-B in aircraft periodically broadcasts general
navigation information via long range RF signal, and anonymous ground users and the other aircraft can utilize it for
situational awareness and self-separation. The major difference with RFID system is the broadcasting message content:
ADS-B messages contain identification and navigation information for the aircraft, which is standardized, such as altitude,
GPS, identification number of aircrafts, etc. ADS-B has been
recently applied to drones [95], [96] for surveying flight information within the target area. Conventional ADS-B systems
are too large for smaller drones, so smaller ADS-B modules
are required. The Ping2020 family [97] by uAvionix is an
off-the-shelf product for drone-level ADS-B, which can be
attached to the drone flight controller (e.g. Pixhawk [98])
and broadcast flight information through the RF channel.
Currently, Ping2020 has somewhat higher price than expected
(US 2000 per Ping2020i [99]), which blocks large-scale
deployment. Low production cost and nation-wide drone registration systems can construct wide area drone identification
infrastructure for continuous and robust identification.
Drone identification phase can be flexibly configured from
drone-to-others authentication to threat analysis. However,
anti-drone system should clarify each logic that determines
whether or not to neutralize drones to cover any type of
drone intrusion. This determination should have firm criteria
from the detection results, national or international regulations, and auxiliary identifiaction tools. Then, according to
the determination, the proper level of neutralization scheme
must be in accordance with the law. As an intermediate step,
identification system must be defined as a determination tools
that provide zone-safe and regulatory-safe solution in given
circumstances.
V. DRONE NEUTRALIZATION
We use the term Drone neutralization as a component of
anti-drone system which refers to operations that suppress
the threatening drones’ movements. We classify the neutralization methods as destructive and non-destructive. This
classification is valid since it not only presents technical difficulty, but also availability within civil regulations.
Destructing the illegal drones are currently prohibited in
many countries, so non-destructive ways are preferred in
VOLUME 9, 2021
several public constitutions. We address in more detail
non-destructive methods, to achieve high utilization of
anti-drone systems in the worst cases.
Mostly, confirmatory methods such as jamming are
preferred to prevent secondary crises (landing/crash and/or
operational failure). Jamming is confirmatory and also nondestructive, but as discussed above, causes temporary communication paralysis across the target area. Thus, recent
approaches attempt to individually disturb the target drones,
considering their operation features. Table 4 lists several common drone neutralization solutions each of which is discussed
in the following subsection.
A. DRONE HIJACKING
The terms hijacking and spoofing are often used interchangeably in anti-drone domain. We clarify meaning of the terms
in this paper for readability. Drone hijacking means that a
defending operator stake control of the target drone regardless
of the methodology. Drone spoofing means that the operator
generates a fake signal to prevent the target drone from moving as intended by the original controller. The main difference
between hijacking and spoofing is post-attack behavior. The
original controller cannot control the drone after hijacking,
whereas spoofing signals can be used to hijack drones.
The reason for this definition is mainly the need for control
deprivation. Usurping the original operators’ control could
include jamming or hacking before the anti-drone system
obtains actual control. Thus, hijacking could be technically
and regulatorily challengeable, but it is more robust then
spoofing after successful deprivation. In any case, both should
be investigated and for confirm defense.
Most drones establish a tightly coupled or paired connection with the operator, and hijacking focuses on breaking
this pairing. Trujano et al. [115] proposed a system to break
the pairing using a jamming signal and instantly re-attach to
the attacker’s controller to take control. Donatti et al. [100]
proposed a drone hijacking system by increasing RF signal amplitude. The authors considered drone control packet
decoding schemes and validated the proposed system with
a prototype. Drone hijacking is an ideal approach in terms
of safe capture or landing, and facilitates follow-up investigation. However, extending coverage and measures, e.g.
autonomous flight, drone communication protocol, etc., are
major challenges.
B. DRONE SPOOFING
Spoofing the drone signal can be used to hijack drones or
confuse their flight routes. Drones generally control their
location and altitude from the operator’s RF signal, but use
sensor outputs to determine current status. GPS signals are
key data to determine current drone position in manual or
autonomous flight. Noh et al. [101] proposed a system to generate fake GPS signals to deceive the drone’s GPS receiver,
and make the drone mistakenly calculate its position. The
authors aimed to secure hijacking related to the GPS failsafe
mode of the drone internal system and quietly send the drone
42643
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
TABLE 4. Drone neutralization technologies.
to a specific location. Simple spoofing techniques that distract
drones can take advantage of several types of sensors, which
may be cooperatively combined. Deceiving drone sensors can
be achieved following a wide range of approaches regardless
of the communication protocol, but in the absence of separate
safety measures for areas outside a certain range, accidents
such as crash landing may occur due to unpredictable drone
operator control.
C. GEOFENCING
Geofence based drone neutralization systems prevents target
drones from approaching a specific point. Various methods
have been studied to block the trespassing drones, including
the techniques discussed in Sections V-A and V-B. However, the most generally adopted and implemented approach
is that the drone self-determines whether or not the drone
lands from its current location [116]–[118]. Geofence technology for drones is classified into two types [118]. Dynamic
geofence propagates information regarding restricted flight
zones, and static geofence uses a flight permission information repository that any drone can access. Most commercial
drones with common flight control stacks, e.g. PX4 [119]
and ArduPilot [120], have internal auto-landing modules for
safety. This method effectively prevents hobby drones from
invading unlicensed areas, but cannot defend a modified or
remodeled drone – disabling automatic landing systems built
into the drone controller. Since the system relies on the
drone’s internal navigation logic, malfunctioning drones may
allow trespassing into the secured area. Further preemptive
geofencing studies are required to address these limitations,
which may utilize spoofing and hijacking techniques.
As discussed in Section I, drone neutralization terminologies are somewhat confused due to the variety of mechanisms
42644
and their consequences. Fig. 6 summarizes representative
non-destructive drone neutralization mechanisms, and Fig. 7
shows their technical relationships. The hijacking, spoofing,
and auto-landing sets (H, S, and G, respectively) refer to
top tier scheme classifications. Each intersection refers to
neutralization scheme collaborations, e.g. [100] for H ∩ S,
and [101] for S − H − G. Thus, anti-drone system designers
can evaluate redundancy of neutralization deployments by
this approach. Anti-drone systems must prepare compositive
systems including S ∪ H ∪ G to cope with highly secure
drones, such as high-level anti-hijacking systems.
D. DRONE JAMMING
Drone jamming focuses on paralyzing radio communication
between the target drone and controller by strongly interfering RF signals, which can be any kind of empty packet
signals within a targeted frequency range. The general purpose is to make the target opponent fall into an uncontrolled
state where they cannot exchange external communication
signals [121]–[123]. Jamming technology can be classified
into various types by the different objectives and coverages.
We introduce some representative classification criteria.
1) The jamming system can be classified into directional [105] or omnidirectional [106] jamming by the
operating direction. The former focuses on a specific
direction, and the latter can jam all direction.
2) Stationary jamming is when the jamming system is
installed at a fixed location, such as a strategic location or base station, whereas mobile jamming is when
the system is operated from portable devices such as
handheld or vehicle mounted [124], [125].
3) Narrow or wide jamming is distinguished by the system
bandwidth [126]–[128].
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
L3 [136]). However, since drones generally do not follow
conventional communication protocols, we will not describe
these approaches in details. Jamming can also be implemented by degrading the target communication quality [137].
However, anti-drone systems aim for complete drone neutralization, hence jammers that only reduce drone performance
were not considered in this paper.
Jamming is a simple, robust, and wide-range solution with
low failure risk, hence most anti-drone systems adopt jamming as their major neutralization scheme [84]. However,
since jamming techniques mainly uses electromagnetic signals, they can have significant unintended impacts, including
TV broadcasts, telecommunication, or even the air traffic
system. Thus, most countries strictly prohibit jamming technology in public. The USA Federal Communications Commission strictly forbids use of any radio jamming system at
consumer level [4], and the UK Office of Communication
also restricts jammers for any purpose interfering with radio
communication [5]. Many countries provide guides to use
jammers, but they have very strict legal constraints. Therefore, there is almost no practical jamming technology possible for civilian users. Thus, non-military grade anti-drone
systems should generally be designed without jammers.
E. KILLER DRONES
FIGURE 6. Typical non-destructive neutralization mechanisms.
We use the term killer drone to mean legal drones that track
target drones and attempt to damage them [43]. To distinguish
killer drone from drone capture, we limit the killer drone
scope to solutions that physically strike invading drones.
Killer drones require reactive and real-time decision making
regarding incoming drones, high accuracy drone flying path
estimation [87], [89], and outstanding physical durability and
mobility [113]. Using drones to damage illegal drones is very
early stage technology, and requires considerable patience
for adoption into commercial anti-drone systems. Swarming
killer drones with distributed intelligence [138] and precise
tracking systems [87], [89] could be a promising solution
for drone fleet multi-faceted attacks. Similar to jamming
and radar, scrambling killer drones is subject to regulatory
restrictions [139], but can be mitigated by policy changes and
technical maturity in drone management systems [140].
F. DRONE CAPTURE
Drone capture approaches physically bind the target drone
with various tools, generally some type of net or similar rather
than military ammunition. We divide drone capture systems
into two groups depending on the capture mechanism.
FIGURE 7. Non-destructive drone neutralization scheme relationships.
4) GPS jamming aims to cause the drone GPS system
to malfunction [129], [130]), whereas communication
jamming aims to interrupt communication between the
drone and controller [131]–[133]).
Exceptionally, there are other jamming approaches targeting specific network layers (e.g. L1 [134], L2 [135] or
VOLUME 9, 2021
Terrestrial capture systems [109], [111] are human-held
or vehicle-mounted, and are available in wide range of
net sizes and numbers of rounds.
• Aerial capture systems [107], [108], [110] are installed
on defender drones, with restraint to the amount and
size of net bullets. Aerial capture provides much more
precision and response speed than terrestrial capture due
to drone mobility, but high requirements for tracking
•
42645
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
accuracy and speed raises the possibility of neutralization failure.
This classification comes from the entire difference in the
implementation phase of each. Terrestrial capture system
should consider the coverage of each device and the optimal
spots for safe capture. Furthermore, improvement strategy is
to increase the effective range of the devices, which has a
higher balance between the cost and the performance than
aerial case. On the other hand, aerial capturing devices have
a higher tradeoff at weight and the performance due to the
limited load of the drones. In addition, aerial capturing system
mainly considers the traceability of the drone itself; the flight
performance and the formation strategy of the drones are key
issues of the system. It is not desirable to determine which
ways are correct, so both approaches should be investigated
in the future.
As drone detection mechanism utilizes the features of
drone operation, drone neutralization also exploits them and
invokes unintended operation. However, not all neutralization
methods always meet the anti-drone system administrators’
goal, and may vary from case to case. Thus, multiple neutralization schemes should be prepared in one package and
increase the successful rate of each. It is also important to
continuously follow up the drone-safe technologies such as
anti-spoofing and anti-hijacking, each of which was originally suggested without malice. In addition, anti-drone system should be able to plan its neutralization, considering the
effective range of the scheme and the estimated flight path,
detailed in Section VII-C.
VI. ANTI-DRONE SYSTEM USE CASES
Recently, most installation examples of officially released
anti-drone systems are in airports and prisons, or temporary
but important meetings. Current security systems tend to
install a drone detection sensor package but are not integrated
with identification and neutralization phases. Rather, they
rely on human guards carrying drone neutralization equipment, hence are vulnerable to rapid and elaborate attacks
from high technology drones. This section introduces some
practical use cases for current anti-drone system installations and addresses remaining requirements for public safety
against illegal drones. Table 5 summarizes the use case
details.3
A. ANTI-DRONE SYSTEMS AT AIRPORTS
After drone incident [13], Gatwick International Airport
installed a military grade anti-drone system [141] using the
British army’s laser based destructive system [112]. Northwest Florida Beaches International Airport is a coastal airport, and installed a detection system to detect both bird and
drone incursions [142]. Copenhagen International Airport is
one of the most population intensive airports in North Europe,
and hence deployed the Mydefense anti-drone system [143],
developed by Denmark. Muscat Airport in Oman installed the
3 We only included confirmed information from authorized media.
42646
German AARTOS anti-drone system [82], [151], characterized by scalability of the defense area up to 50 km to meet
large airport requirements.
Airports are densely populated facilities with high injury
risk in the event of an airplane accident, hence national security agencies tend to deploy military grade defense systems.
Anti-drone systems are expected to increase gradually as
international airports in each country develop further from
their current pilot systems. However, radars are widely used
for air traffic management (ATM), so many airport cases are
limited to non-radar components. It is essential to deploy sufficient alternative solutions to ensure safety against unauthorized drones, including vision and sound based approaches.
B. ANTI-DRONE SYSTEMS IN OTHER FACILITIES
Various non-airport anti-drone systems have been implemented against illegal drone intrusion. Suffolk Prison
deployed a drone detection system to prevent smuggling
prohibited items, such as drugs and mobile phones [144].
New York’s Mets City Field installed a drone detection system to protect against unlicensed broadcast
of the matches [145]. Zhejiang University independently developed and deployed anti-drone system named
ADS-ZJU, with sensor fusion and automatic jamming technology [84]. PyeongChang Olympics Stadium deployed
interceptor drones to capture illegal drones [146], and
temporary drone detection systems were deployed the at
University of Nevada, Las Vegas (UNLV) during US presidential debates [147], Davos during the World Economic
Forum [148], and Buenos Aires during the G20 summit [149].
Israel has developed the Drone Dome anti-drone system [150]
covering the whole country.
C. REMARKS
The use cases considered here indicate that anti-drone systems are currently installed where real drone threats, such
as smuggling or terrorism, are expected. This implies that
anti-drone system design remains introductory and relatively primitive compared with actual drone incident urgency.
A generalized and diverse strategy for anti-drone system
design is urgently required. Considering the high mobility and accessibility of drones, anti-drone systems must be
installed on large scales to prevent sporadic drone incidents.
However, existing anti-drone systems are generally military
grade and only installed in important facilities, hence other
sites are vulnerable to drone attacks and may need appropriate anti-drone systems according to national regulations.
Furthermore, majority of anti-drone systems include only
drone detection and alert systems with identification and neutralization stages are performed by people, mostly soldiers.
Integrating detection, identification, neutralization schemes,
and automating the overall system would greatly improve
anti-drone system accessibility and reduce the labor costs.
Section VII proposes some guidelines for efficient anti-drone
system design.
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
TABLE 5. Anti-drone system use cases.
VII. ANTI-DRONE SYSTEM GUIDELINES
Comparing with military installations, non-military facilities
are significantly disadvantageous to defend against illegal
drones legally and technically. Rapid advances in imbedded
systems make drones continually get smaller and faster in
obstacle-rich 3D space [69], [152], with enormously increasing payload capability. Most countries regulate drone use
for industrial and market applications, and require urgent
implementation of a systematic and rigorous drone defense
system at major facilities. Several drone attacks were reported
in the 2010s, not only for military bases [13]–[16], [18],
[20]. Non-military drone intrusion can cause considerable
economic damage, but applying ideal anti-drone systems in
overall area is almost impossible not only due to budget constraints, but also workforce. Furthermore, excessive response
to hobby drones can impose considerable capital redundancy,
and still may be vulnerable to subsequent serious attacks.
Utilizing the surveys in Section III–V, this section proposes a guideline for designing non-military anti-drone systems, including where to deploy the equipment, what method
should be chosen at neutralization, and how to define integrated response procedures.
Given the scope of this paper, we do not consider amendments to legislative provisions regarding drone and defense
system permits, etc.
A. DETECTOR DEPLOYMENT
An ideal drone detection system could form a comprehensive
deployment of high-performance devices with high density,
but system management and installation must be considered
to operate a cost-efficient drone defense system. The system
must also be able to intensively monitor critical points where
catastrophic accidents such as massive explosions or top
secret leaks could occur. Furthermore, increased detection
accuracy is essential to obtain as much information as possible by combining multiple detection methods (Section III-E).
We propose a superpositioning strategy for drone detectors
considering relative importance of the areas. First, we categorize detection methods by quantitative and qualitative
VOLUME 9, 2021
features, and show deployment examples for an airport and
industrial facility.
1) DETECTION EQUIPMENT CATEGORIZATION
Table 6 shows the categorization scheme that we proposed.
We considered not only the technical mechanism of detection, but also detailed specifications related to detection performance, such as operating angle and installation method.
The main objective is to improve the detectability of high
priority areas; hence we set the criteria that differentiates
existing detection systems. For example, radar based detection has technical specifications at product level, such as
directional/omnidirectional, stationary/non-stationary, detection range, and whether the drone is identified.4 On the other
hand, if a detection system combines multiple sensors (e.g.
vision+acoustic, Section III-E), new type should be created
to reflect differentiated detection performances. Categorizing available equipment according to features rather than
methodology means the system can deploy detection systems
in terms of various metrics, which lays the foundation of
detection system abstraction.
2) AREA PRIORITY CLASSIFICATION
Understanding the features and characteristics of defense
area is essential for optimal drone detection deployment.
The proposed approach analyzes defense area spatial usage
and classifies area priority into several levels. Classification
criteria reflects the main purpose of defense against drones,
which generally effects civilian and major property safety.
Figs 8 and 8b describe a fictitious airport and industrial
plant, respectively, as classification examples.
We mainly considered safety risks for airport passengers
in the airport classification. The most threatening airport
situations would be the one that drones cause aircraft crash
at takeoff or landing, which could result in massive human
casualties [153]. A relatively small drone can harm a flying
4 There are technical differences between the radars that can detect a
drone-size object (low RCS) and identify the drone [35], [69], [70].
42647
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
TABLE 6. Detection equipment categorization criteria.
airplane if it enters the jet engine or damages the wings, particularly during take-off or landing. To prevent this, the runway and surrounding surfaces must be closely monitored,
and any abnormal condition quickly propagated to control
aircraft takeoff and landing. We exploited International Civil
Aviation Organization (ICAO) regulations regarding obstacle
limitation surfaces [154] to prioritize areas within the airport.
• Class I included airplane gliding and landing areas with
altitude range to drone-identifiable altitude, for the highest priority of protection.
• Class II included navigation safety facilities and population intensive areas that could result in massive human
injury.
• Class III included areas that could have short and long
term impact on airport operation and ATC in the event
of attack. This considers social and economic risks of
drone-aircraft collisions, which could temporarily paralyze ATC facilities, or cause airport shutdown.
• Class IV included areas protected by automatic drone
guidance systems such as Geofencing (Section V-C),
and boundaries where drones should turn back, such as
conical and horizontal airport surfaces.
There are no global regulations relevant to area designation
for the industrial factories. Therefore, we considered both
safety and security of the facility.
• Class I included areas with the highest potential for
large-scale incidents, such as explosion from drone
crash.
• Class II included population intensive areas, similar to
the case for the airport.
• Class III included security sensitive areas to prevent
drones from breaching confidentiality.
• Class IV included remaining and detectable outer
perimeter areas, to proactively detect and respond to
drone intrusions.
Fig. 8 shows priority assignments for the airport and
factory examples. Various areas were scaled for visibility;
Class IV space was actually much larger than the sum of
Class I to III. The system can indicate where to focus detection resources to minimizes potential damage caused by a
drone incident. Note that lower class areas do not mean that
these areas represent a weak point. The main goal of this
classification was not to reduce the likelihood of detection
42648
in low-risk areas, but to effectively expand the detection
network in cost restraint.
3) ABSTRACT FORMATION
From detection categorization and priority classification,
we show example drone detection deployments using the
abstracted detection equipment. Fig. 9 shows deployments
for the airport and industrial facility cases. Figs. 9a and 9b
show two-dimensional views of the sample areas (left side),
indicating area classes, for visibility and better structural
understanding. The right side of the figures show results for
3 types of stationary equipment arranged in the defense area.
Type 3 equipment has wide surveillance range and hence
is installed at the center of the defense area to include as
many essential areas as possible; whereas type 2 equipment
is directional, hence devices were placed on entry surfaces
on both sides of the runway, and two others on the runway
and airport terminals considering the arrangement of Class III
facilities.
Real installations will have many more things to consider
when determining physical location of the equipment, such as
radio frequency bands, spatial margins with wireless equipment in the target area, etc. Representative considerations are
as follows.
Radio wave environment. In the case of drone radar,
it is necessary to investigate the risk of reducing detection rate due to radio interference between the drone
radar equipment and nearby existing radars, and check
frequency bands employed. Most countries set regulations on installing new radar sites [155], in terms
of frequency band and spatial distancing, hence prior
consultation with relevant institutions (e.g. US Federal
Communications Commission (FCC)) is required.
• Legal operator boundaries. National regulations
related to anti-drone solutions (particularly radar, camera, RF jamming, and killer drones) should be considered prior to deployment. The system designer should
also check legal right to arrest drone owners and public regulations regarding destroying or damaging the
drones, to ensure system installation cost is more effective while accomplishing the security requirements.
• Physical environments. Fig. 9 shows that existing
facilities can block radar scanning, vision, etc. The
•
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
FIGURE 8. Priority classification examples according to drone threat level.
FIGURE 9. Detection system deployment scenario.
anti-drone system designer should carefully locate
selected detection devices considering building and surrounding terrain profiles. Acoustic detection devices
should not be placed around noisy environments, and
wired or wireless networks should be optimally configured to collect results from multiple detection systems.
• Radio interference. RF interference with existing radio
based systems (weather or aviation radar, ATC components, etc.) must be concerned before installing the
drone detection radar, which could detrimentally impact
overall RF equipment performance. Discussions regarding interference between 5G access networks and drone
radar [73] should also proceed nationwide.
The main objective for the proposed deployment was sustainable design for drone detection devices’ configuration,
which could be semi-permanently employed by abstraction.
The deployment process could be applied when new detection
equipment was introduced or protection priorities changed.
Future work on this scheme should be an autonomous
algorithm for selecting and deploying the detection devices
considering installation cost and system features. Optimal
placement of drone detection network can be obtained with
VOLUME 9, 2021
the proper objective function and efficient algorithm design,
which greatly improves security against illegal drone incursion within a given budget.
4) AIRSPACE PRIORITY CLASSIFICATION
Civilian hobby drones can achieve over 3 km maximum
height [156], by the lack of the awareness of drone regulation. Area priority classification effectively protects a
specific target area, but it is necessary to classify large
areas by priority to expand protection range to the national
airspace and prevent accidents of various aircraft types, such
as PAVs. Thus, we suggest an airspace priority classification, based on ICAO regulations [157]. Wide-spread concerns
about drones in operating airspaces have raised the awareness for suitable drone regulation [158]–[160]. Airspace
is defined across a wide area and it is difficult to cover
with conventional local anti-drone systems, hence large-scale
airspace drone defense system should be constructed
with global (or national) drone identification infrastructure (Section IV), containing long-range identification and
detection equipment.
Fig. 10 shows space priority classification according
to ICAO airspace classification criteria. Airspaces above
42649
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
FIGURE 10. Airspace priority classification.
airports (B to D) are usually shaped as a stack of cylinders
with larger radius at higher altitude, reflecting airplane flight
and landing paths. However, drones draw landing and takeoff
trajectories relatively freely, so we allocated a large cylindrical area near the airspaces. We designated Class I areas
to large and busy airports (B and C), and Class II areas to
relatively small airports (D). Airspace A includes airplane
flightpaths, and hence we allocated these to Class II since
drone collision at this airspace could cause forced landing or
worse. We set airspace E to Class III due to human accident
risk, and airspace G to Class IV for preemptive drone monitoring.
The hierarchical configuration with local anti-drone systems such as in airports and factories will provide a global
view of widespread provocation of drones, and enable cooperative tracking and neutralization. Every country has different airspace regulations [161], and some countries do not use
some of their airspaces, so the system designer should reflect
nationwide potential threats and geographical features.
B. THREAT LEVEL ASSESSMENT
When a drone illegally approaches, it is mostly unavailable
to identify who or what the drone belongs to or it intends
to do. Thus, anti-drone system should be able to assess the
potential threat of unknown drones from the defender’s point
of view. We can obtain an exemplary threat model from drone
detection results as a numerical value R. We can determine R
with
R = (Robject + Rpath )Rtime ,
(1)
where Robject , Rpath , and Rtime refer to threat levels induced
by the drone, drone flight path, and remaining response time,
respectively. We set maximum Robject and Rpath = α and β
(constants), respectively, and Rtime = 1.
In our model, R quantifies the likely damage that would
occurs if the observed drone actually attacked in a certain
area, considering the currently measured factors. Robject represents the physical impact from the drone crash, and Rpath
42650
the estimated damage for a given crash point. Powering factor Rtime indicates how much the damage realized, and R
becomes the realized damage of the observation when Rtime
is maximized to 1.
Eq. (1) can be made more flexible by parameterization. For
example, Robject could be set to a constant if the detection
network only provides the drone position. Since derivation of
Robject requires detailed drone information, constant Robject
can produce wide R range through the remaining parameters Rpath and Rtime . The derivation of each parameter is as
follows.
1) THREAT LEVEL BY OBJECT
Ideally, the drone’s physical threat level is defined by its
physical properties such as kinetic energy, but the collected
information can be limited. Thus, we model the derivation
of Robject with complex conditions. Basically, kinetic energy
is determined by drone mass or weight (maximum takeoff
weight) and speed, where K = 12 mv2 . Detecting drones with
RF scanners can provide detailed information, such as model
name and weight, from a predefined database. This allows us
to infer the drone’s bare bone weight and potential payload,
providing an estimate for increased threat level if detection
device catch the payload of explosives.
Table 7 shows an example physical threat level classification scheme considering drone characteristics that can
be obtained through the detection systems. We referenced
criteria for classifying threat level from the kinetic energy
at [162] and noise at [163]. Most factors can be determined
once the drone’s commercial model is identified, including
the mass, and the threat level can be more accurately derived.
Otherwise, the system approximates drone weight from the
size and default density from the database. Low noise drones
are classified as higher threat since noise can help humans
track and evacuate drones easily, and reduce the damage.
Table 8 describes how to calculate the threat level according to drone physical characteristics, where values were
obtained from Table 7, and Ndrone refers to the number of
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
TABLE 7. Drone classification by physical features.
TABLE 8. Robject derivation.
swarming drones. This score is not an absolute range for
Robject , and the system administrator can change the weights
on demand. Partial threat factors from size, energy, scanability, and loaded objects are calculated and then summed to
obtain the object-wise threat level. The overall threat level of a
drone swarm is derived by multiplying the object-wise threat
level to Ndrone . If the drone is not detected by RF scanner or
the system experiences difficulty in accurate determination of
some parameters, the safest option is to conservatively set any
uncertain parameter to its maximum value to induce strong
response.
Examples of Robject . Let
Wkinetic = Wnoise = Wloaded = Wscannable ,
and two DJI Phantom 4 drones weighing 1.3 kilogram are
approach at 72 km/h, 70 dB noise, without additional transport. Then Robject ≈ 0.134α, which is relatively low. However, if the drone approaches major facilities, such as runways
or civilian-intensive spots, the comprehensive threat level is
set higher. Robject then used to calculate the total threat level
α + β in addition to Rpath and Rtime .
Similarly, if an unknown drone with approximate size
800 × 700 × 400 mm approaches at 60 km/h, 80 dB noise,
with a lightweight weapon, then Robject ≈ 0.453α. This case
is considerably higher than for the two DJI Phantom 4 drones,
mainly due to loaded weapon and lack of scannability. This
case of threat level is valid since the drone may intend a
terrorist attack, and be unable to neutralize by geofence or
hijacking.
2) THREAT LEVEL BY FLIGHT PATH
We adopt the defense area analysis of Section VII-A to
determine drone threat level in various cases. In example
of airport, threat level is low if the drone is flying around
airport conical space, and the system can observe automatic
landing behavior or prepare delayed but safe neutralization
VOLUME 9, 2021
methods, e.g. capture. However, if the drone approaches near
the runway, where it could impose severe damage, the system determines the increased threat level and can prepare
immediate, confirmatory, risky, or expensive neutralization
methods, e.g. jamming or firing, as appropriate. Multiple
neutralization methods are required to efficiently cope with
various drone attack situations. If risky options, e.g. jamming,
are not permitted, then the system should have an emergency
hotline to related agencies, such local police or military bases.
Based on area priority classification, we determine potential threat levels for each area. Table 9 shows an example
derivation for each class’s potential threat, Ti , where i =
1 . . . 4. First, the system derives a numerical value for each
drone incident, and checks if the incident can happen in the
each area. Then it calculates the sum of incident values of
each area, and divides by the largest sum to obtain coefficients
for Ti , where T1 = 1, and finally multiply each coefficient by
β to derive Ti . Thus, we quantify the expected damage from
a successful drone attack for each area class.
Threat level for the drone’s flight path can be determined
from the potential threat levels for each area, applying higher
levels when the drone path includes high-priority areas. Let
−
→
F be the aggregated drone flight direction (unit vector),
−→
measured from tracking data over some period, and Fd,i be
the unit vector from its current location to the nearest Class i
−
→
−→
point, as shown in Fig. 11. F and Fd,i can be calculated from
drone tracking data and the drone threat level Rpath is
X −→ −
→
Rpath = Tc +
(Fd,i · F )wd Ti ,
(2)
i∈S ,i6 =c
where S is the set of area indices; c is the section containing
the drone; and
1
wd =
,
(3)
1 + exp (d − D2 )
where d is the distance between the drone and area of interest
and D is the system’s maximum detection range. wd indicates
42651
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
TABLE 9. Potential threat derivation according to priority classification (airport).
FIGURE 11. Rpath determination for illegal drone flight direction.
a weight of threat with respect to a specific area, which
increases when the drone approaches to the area. In sum, Rpath
increases if the drone approaches high priority areas, hence
the proposed system can differentiate threat levels between
(for example) a drone circling the airport conical space and
one rushing toward a runway.
3) THREAT LEVEL BY AVAILABLE TIME
Response time affects anti-drone neutralization method selection. We employ time-wise threat as a weight to enable the
system to respond to emergency situations. However, estimating available response time is challenging due to uncertainty
regarding the illegal drone’s purpose. Therefore, we modeled
Rtime assuming the worst case of active drone attack from
factors collected by the drone detection and tracking system.
Rtime can be expressed as


tavg
Rtime = min
,
1
,
(4)
−→
Dcritical ÷ |−
v−
drone |
where tavg is the average response time for the available
neutralization methods; Dcritical is the minimum distance to
a critical point; and vdrone is the average drone velocity. Rtime
is generally larger than the expected time to prevent tardy
system response. Critical area can be assigned by using the
proposed area classification method (Section VII-A2) or any
other useful system. The denominator in (4) expresses the
remaining response time, hence the system should deploy
fast and effective neutralization methods as Rtime → 1.
Similar to Rpath , Rtime should be updated periodically as drone
42652
FIGURE 12. Risk model for drone neutralization.
neutralization progresses to immediately react to changing
situations.
C. RISK MANAGEMENT
With few exceptions, neutralization carry the possibility for
further damage. For example, a promising neutralization
technology is the drone capturing gun [109], [111], which
fires a large net bullet to wrap the flying drone. However, successful neutralization means the captured drone immediately
fall to the ground, potentially causing additional damages
such as explosion or human injuries. Thus, it is essential to
decide where and when to neutralize the incoming drone,
considering the risk from drone neutralization. We propose an
approach to search an optimal position to intercept the drone
after the threat assessment in Section VII-B.
To derive the risk model for drone neutralization, we consider a simple scenario employing the drone capturing gun,
as shown in Fig. 12. Suppose a drone is flying across the
−
→
defense area, with current location dt and estimated flight
−→
−
→
path e(t). The selected neutralization device is located at kt ,
and moves toward the estimated drone route with speed v.
→
The system intends to intercept the drone at position −
p , and
estimated response time tresp is
−
→
→
|−
p − kt |
.
(5)
tresp =
v
−−−→
The capture gun fires the net at e(tresp ), hitting the drone if
it is within the device’s effective range Deff , and the drone
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
FIGURE 13. Estimated risk from drone flight path and neutralization
point.
−
→
subsequently falls to the ground at some crash point c( dt ),
a probability distribution of coordinates. The expected total
→
risk r(−
p ) can be derived from the associated potential threat
level for crash or collision for each possible crash coordinate,
derived from the spatial threat assessment (Section VII-B2),
I
−
→
→
→
r( p ) =
S(−
c )pcrash (−
c ),
(6)
C
→
where C is the set of possible crash points; S(−
c ) is the
−
→
−
→
potential threat for location c ; and pcrash ( c ) is the crash
→
probability for −
c.
−
→
Thus, c( dt ) and Deff must be already known to derive an
estimated response risk. Smaller crash point clusters imply
→
larger area of low r(−
p ) values, and larger Deff implies larger
−
→
available response area. c( dt ) and Deff are strongly related to
the particular neutralization system selected, and hence are
essential parameters to evaluate neutralization performance.
Fig. 13 shows a simple drone flight simulation coded in
MATLAB to verify the proposed risk model validity. We
formed a (100, 100, 100) simulation space, with estimated
flight path from (5, 5) to (100, 90) and 3 designated areas with
potential threats levels {50, 30, 20}. The drone response team
−
→
(device and carrier) were initially located at kt = (80, 80)
−
→
and r( p ) distribution was represented as a three-dimensional
→
mesh. As shown, r(−
p ) is a dynamic value depending on
selected neutralization points. In particular, the neutralization
method was unavailable for some of the nominally available
area due to range limitation. Then, the optimal operation point
can be approximately (60, 80) with lowest risk. Designing
risk model can determine where to deploy the neutralization
process with lowest risk in dynamic situations.
VIII. ADVANCES IN DRONE TECHNOLOGY
Before the anti-drone system get spotlighted as the security
solution for drone incidents, drone researchers rather studied
the security solutions for the drones to defend against the
malicious attacks [164]. Drones are now widely acknowledged as potential weapons, and anti-drone technologies
have been widely studied to defend against malicious drone
VOLUME 9, 2021
FIGURE 14. Advances and requirements for detection technologies.
attacks. However, anti-drone and drone-safety domains are
complementary, similar to hacking and network security. This
section discusses drone-side defensive technologies and propose future directions for anti-drone systems.
A. ANTI-DRONE NULLIFICATION TECHNOLOGIES
Drone defense solutions have evolved in part to avoid risks
from expanded industrial drone use. Increased drone automation and vastly improved stability have greatly improved
drone security, and hence many current anti-drone solutions
can fail to defend against modern drones. We consider several
drone safety systems that could potentially cause anti-drone
system failure in terms of detection and neutralization. We
replace the concerns about drone identification technology
with [165], which solves the existing problems of remote
drone registration systems.
1) DRONE DETECTION AVOIDANCE
Disturbing the detection sensor or minimizing drone frame
features are the main detection avoidance approaches, also
called drone stealth techniques [166]–[168]. Most stealth
techniques focus on lowering the frame RCS. Already, some
micro-UAVs (e.g. CrazyFlies [169]) are effectively undetectable at sufficient range to allow eavesdropping or spying
on confidential facilities, and if the drone is equipped with
encryption modules, RF scanners may not detect its approach.
A fleet of drones can construct a high security network
using advanced micro-computers and lightweighted security
schemes [170], [171] that would be impossible to crack until
the drones completed their invasion. Indeed, RF scanners
may become obsolete unless cracking technology catches up
with encryption schemes or adopt working quantum computing devices. Meanwhile, Oh et al. [167] proposed a noiseless drone design to avoid acoustic detection, and particular drone shapes can significantly reduce camera detection
accuracy [172], [173] since vision systems identify drones by
shape.
Drone detection technology struggles with rapid drone
evolution in terms of size, speed, shape, and noise. Fig. 14
42653
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
compares the advances in stealth technologies future requirements for detection technology, based on detection accuracy and range from Section III, required performance from
Section II, and future objectives for drone technologies. Current detection solutions grant limited safety against invading
drones. Rather, cooperation with dense and large-scale drone
identification networks may be better approach to determine
if the flying object is harmful or not. It means, detection system should observe wide types of flying suspicious objects,
regardless of feature, and let identification system decide
to neutralize each. If so, then drone detection can act as a
tracking system, another essential role in drone response.
2) DRONE NEUTRALIZATION AVOIDANCE
There are many drone neutralization techniques (Section V),
and hence safety solutions also vary. Anti-jamming is the
complementary solution for jamming systems. Drone jamming is currently a major neutralization choice despite of
the high risk and strict usage regulations due to simplicity, immediateness, versatility, and wide range. Although
most drones can fly autonomously, breaking the connection
between invading drone and its operator may prevent illegal
attempts such as confidential information acquirement. However, jamming the licensed drone bands may be ineffective
against modified drones using other bands [174] or multiple
bands [175]. Although wideband jamming appears a clear and
robust solution, operational cost and risk should be carefully
considered along with regulatory restrictions.
Non-destructive methods (Section V) tend to assume
a detected drone’s operation method, such as what protocols employed and whether a drone is automated.
Hijacking only works for manually controlled drones and is
ineffective against self-navigating drones. Similarly, although
spoofing can misdirect or kidnap target drones, vision based
navigation [176], [177] may avoid GPS-spoofing. Destructive methods, such as killer drones or drone capture, have
wide potential to neutralize the drones but may struggle
with high-speed obstacle avoidance capabilities [178]–[180].
Thus, anti-drone systems should derive universal, robust, and
precise strategies for drone response scenario, as discussed in
Section VII-C.
B. ANTI-DRONE SYSTEM ADVANCES
Current anti-drone systems are under pressure to establishment safety and security against drones, which remains challenging. Subsequent sections list constructive approaches for
anti-drone systems to achieve drone defense, from global
philosophy to specific methodologies. Since the proposed
statements require longer time than expected because of
the regulatory issues instead of technological difficulties,
national or world-wide discussions also be actively processed
to avoid the global threat of drone incidents.
1) SYSTEM STANDARDIZATION
As discussed above, technical competition between drone and
anti-drone industries causes rapid new product developments
42654
while beating the opposite side of systems. Thus, viable
anti-drone systems must be capable of continuous updates,
which means components must be easily replaceable and
compatible with sustainable architecture. Anti-drone system
component standardization is essential, such as a form of
high-level architecture [181], to allow advanced component
designs to be quickly evaluated and adopted in the empirical
environment.
2) SAFE CHANNEL IN JAMMING
Because of the technical difficulties in other neutralization
methods, jamming is still the last bastion of the anti-drone
system. To reduce the risk of the use and maintain a network of anti-drone systems and target facilities, the available
channel for the defenders, named Safe channel, should be
required. The safe channel can be designed by the ultra-low
band RF or the other mediums, such as visible light [182] or
acoustic signal [183].
3) LARGE-SCALE DRONE MANAGEMENT
Most countries have drone regulations [139], but drone incidents still occur. Thus, fine-grained and strong policies are
required for drone defense. Currently, anti-drone systems
have difficulties identifying and responding to illegal or intrusive drones, so strict drone management regulations, such as
installing hardware identification devices on drones, could
help reduce the burden on the system.
IX. SUMMARY AND CONCLUSION
This paper discussed non-military grade anti-drone systems.
Our findings can be summarized as follows.
•
Drone detection. Modern detection solutions guarantee
a certain level of drone detection accuracy by integrating
multiple detection systems. Each methodology has performance limitations in terms of detection range, functionality, weather dependency, etc., so the anti-drone
industry tends to construct hybrid detection systems.
However, administrators should analysis the defense
area to design optimal detection systems and improve
drone detection efficiency. From the survey, we suggested a guideline for installing anti-drone detection system considering efficiency and priority. The proposed
guidelines include abstract classifications for detection
equipment, priority classification for defended areas,
and actual system deployment examples for airports,
industrial facilities, and airspaces. Detection system
should be tightly coupled with fine grained drone identification networks to provide viable drone tracking and
neutralization solution.
To sum up, considering current performance of detection technology and capability of the drones, each
mechanism should be improved in terms of range and
accuracy to track the drones with advanced stealth functions. In addition, sensor fusion technology must compensate for the flaws in each method while increasing
VOLUME 9, 2021
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
cost efficiency. Meanwhile, anti-drone system designers
should make their own layout of the detection system
and map it to the actual equipment that matches their
requirements. This coexistence of designers and developers can lead to the technical advances in the drone
detection system.
• Drone identification. Empirical adoption of drone identification systems is earlier stage than detection and
neutralization systems due to requiring regulatory cooperation, such as drone registration policies. Attaching
active transponders to drones, similar to conventional
airplanes, is currently under consideration. Drone identification networks will become more important than
detection alone to overcome evolving drone technologies. Airspace provisions can be further subdivided to
prepare for the emerging PAV industry, and anti-drone
systems must be phased in to defend against legitimate
aircraft.
In short, drone identification technology will play an
important role in future anti-drone systems. Considering the rapid evolution in drone technology and the
high attention in anti-drone, drone identification system
should clearly determine whether or not to neutralize
the observed aircraft. False positive or false negative
of identification can result in the entire failure of the
defense. Proper amendments in drone regulation such as
identification tag are essential to construct reliable identification system. With authentication and regulations,
we claim that drone identification can be more specific
like detection/neutralization procedures.
• Drone neutralization. Drone neutralization schemes
exploit various drone features, including flight mechanisms and communication systems. Neutralization
methods can be mainly classified as destructive or
non-destructive. However, most non-destructive methods may be quickly obsolete due to robust drone security and navigation solutions. Although drone jamming
remains the most popular choice for current systems,
its inherent aggressiveness and anti-jamming developments strongly suggest the necessity for alternative
approaches. Geofencing for drones may prevent unintended accidents from legitimately authorized drones,
but deliberate attacks may need to be defended physically, e.g. killer drones or drone capture.
In summary, anti-drone systems should include multiple
neutralization solutions and utilize them appropriately
to improve defense reliability. Specially, destructive and
non-destructive methods should be separately treated
in system design, and must be carefully selected. Our
guideline that assesses drone threat level from relevant
measured parameters and subsequently derives safe neutralization scenarios could be an abstracted procedure
for future anti-drone systems.
Designing anti-drone systems without incorporating military grade weapons and complying to national regulations remains early stage and exposes vulnerability to drone
VOLUME 9, 2021
incidents. Current anti-drone systems have well-formed
detection, identification, and neutralization stages, but more
accurate and effective systems are required to cope with
high-speed, high-security, and three-dimensional attacks. We
proposed guidelines for designing versatile, available, and
sustainable anti-drone to defend against various drone attack
scenarios. Our proposals address the direction to resolve
technical and structural difficulties for anti-drone system
design, and cope with the advances in drones’ defense
mechanism. We expect this anti-drone system survey contributes to expanding the drone-safety zones without requiring weaponry.
REFERENCES
[1] D. Floreano and R. J. Wood, ‘‘Science, technology and the future of small
autonomous drones,’’ Nature, vol. 521, no. 7553, pp. 460–466, May 2015.
[2] M. Ritchie, F. Fioranelli, and H. Borrion, ‘‘Micro UAV crime prevention: Can we help princess Leia?’’ in Crime Prevention 21st Century.
New York, NY, USA: Springer, 2017, pp. 359–376.
[3] B. R. Van Voorst, ‘‘Counter drone system,’’ U.S. Patent 15 443 143,
Sep. 14, 2017.
[4] FCC Enforcement Advisory, Cell Jammers, GPS Jammers, and Other
Jamming Devices, document FCC RCD 1329(2), Washington, DC, USA,
Feb. 2011.
[5] UK Public General Acts, Wireless Telegraphy ACT 2006, U Legislation,
London, U.K., 2006, sec. 68.
[6] Y. Shapir, ‘‘Lessons from the iron dome,’’ Mil. Strategic Affairs, vol. 5,
no. 1, pp. 81–94, 2013.
[7] P. Wellig, P. Speirs, C. Schuepbach, R. Oechslin, M. Renker, U. Boeniger,
and H. Pratisto, ‘‘Radar systems and challenges for C-UAV,’’ in Proc. 19th
Int. Radar Symp. (IRS), Jun. 2018, pp. 1–8.
[8] A. Chadwick, ‘‘Micro-drone detection using software-defined 3G passive
radar,’’ in Proc. Int. Conf. Radar Syst., 2017, pp. 1–6.
[9] B. Nuss, L. Sit, M. Fennel, J. Mayer, T. Mahler, and T. Zwick, ‘‘MIMO
OFDM radar system for drone detection,’’ in Proc. 18th Int. Radar Symp.
(IRS), Jun. 2017, pp. 1–9.
[10] H. Mazar, Radio Spectrum Management: Policies, Regulations and Techniques. Hoboken, NJ, USA: Wiley, 2016.
[11] A. R. Wagoner, D. K. Schrader, and E. T. Matson, ‘‘Towards a visionbased targeting system for counter unmanned aerial systems (CUAS),’’ in
Proc. IEEE Int. Conf. Comput. Intell. Virtual Environ. Meas. Syst. Appl.
(CIVEMSA), Jun. 2017, pp. 237–242.
[12] P. Nguyen, M. Ravindranatha, A. Nguyen, R. Han, and T. Vu, ‘‘Investigating cost-effective RF-based detection of drones,’’ in Proc. 2nd Workshop
Micro Aerial Vehicle Netw., Syst., Appl. Civilian Use, 2016, pp. 17–22.
[13] BBC. (2019). Gatwick Airport Drone Attack: Police Have No
Lines Inquiry. Accessed: Sep. 27, 2019. [Online]. Available:
https://www.bbc.com/news/uk-england-sussex-49846450
[14] The Local. (2019). 143 Flights Cancelled at Frankfurt Airport Due
to Drone Sighting. Accessed: May 9, 2019. [Online]. Available:
https://www.thelocal.de/20190509/disruption-after-frankfurt-airporthalts-flights-due-to-drone-sighting
[15] U.S. Attorney’s Office. (2011). Massachusetts Man Charged With
Plotting Attack on Pentagon and U.S. Capitol and Attempting
to Provide Material Support to a Foreign Terrorist Organization.
Accessed: Sep. 28, 2011. [Online]. Available: https://archives.fbi.gov/
archives/boston/press-releases/2011/massachusetts-man-charged-withplotting-attack-on-pentagon-and-u.s.-capitol-and-attempting-toprovide-material-support-to-a-foreign-terrorist-organization
[16] BBC. (2018). Syria War: Russia Thwarts Drone Attack on
Hmeimim Airbase. Accessed: Jan. 7, 2018. [Online]. Available:
https://www.bbc.com/news/world-europe-42595184
[17] B. Hubbard, P. Karasz, and S. Reed, ‘‘Two major Saudi oil installations hit by drone strike, and US blames Iran,’’ The New York Times,
Sep. 14, 2019.
[18] C. W. Ripley. (2015). Drone With Radioactive Material Found on
Japanese Prime Minister’s Roof. Accessed: Apr. 22, 2015. [Online].
Available:
https://edition.cnn.com/2015/04/22/asia/japan-primeminister-rooftop-drone/index.html
42655
S. Park et al.: Survey on Anti-Drone Systems: Components, Designs, and Challenges
[19] T. Gibbons-Neff. (2016). ISIS Used an Armed Drone to Kill Two
Kurdish Fighters and Wound French Troops, Report Says. Accessed:
Oct. 11, 2016. [Online]. Available: https://www.washingtonpost.com/
news/checkpoint/wp/2016/10/11/isis-used-an-armed-drone-to-kill-twokurdish-fighters-and-wound-french-troops-report-says/
[20] BBC. (2018). Venezuela President Maduro Survives Drone
Assassination Attempt. Accessed: Aug. 5, 2018. [Online]. Available:
https://www.bbc.com/news/world-latin-america-45073385
[21] G. Ding, Q. Wu, L. Zhang, Y. Lin, T. A. Tsiftsis, and Y.-D. Yao, ‘‘An amateur drone surveillance system based on the cognitive Internet of Things,’’
IEEE Commun. Mag., vol. 56, no. 1, pp. 29–35, Jan. 2018.
[22] P. Andraši, T. Radišić, M. Muštra, and J. Ivošević, ‘‘Night-time detection
of UAVs using thermal infrared camera,’’ Transp. Res. Procedia, vol. 28,
pp. 183–190, Jan. 2017.
[23] HGH Infrared Systems. (2020). HGH Spynel. Accessed: Jul. 28, 2020.
[Online]. Available: https://www.messe-essen-digitalmedia.de/uploads/
E302/pdf/company/hgh-infrared-systems-f9d3f-info.pdf
[24] C. Aker and S. Kalkan, ‘‘Using deep networks for drone detection,’’ in
Proc. 14th IEEE Int. Conf. Adv. Video Signal Based Surveill. (AVSS),
Aug. 2017, pp. 1–6.
[25] M. Saqib, S. D. Khan, N. Sharma, and M. Blumenstein, ‘‘A study on
detecting drones using deep convolutional neural networks,’’ in Proc. 14th
IEEE Int. Conf. Adv. Video Signal Based Surveill. (AVSS), Aug. 2017,
pp. 1–5.
[26] I. Guvenc, F. Koohifar, S. Singh, M. L. Sichitiu, and D. Matolak, ‘‘Detection, tracking, and interdiction for amateur drones,’’ IEEE Commun.
Mag., vol. 56, no. 4, pp. 75–81, Apr. 2018.
[27] S. S. Trundle and A. J. Slavin, ‘‘Drone detection systems,’’ U.S.
Patent 15 282 216, Mar. 30, 2017.
[28] M. F. Al-Sa’d, A. Al-Ali, A. Mohamed, T. Khattab, and A. Erbad, ‘‘RFbased drone detection and identification using deep learning approaches:
An initiative towards a large open source drone database,’’ Future Gener.
Comput. Syst., vol. 100, pp. 86–97, Nov. 2019.
[29] CRFS. (2020). DroneDefense. Accessed: Jul. 28, 2020. [Online]. Available: https://pages.crfs.com/hubfs/CR-002800-GD-2-DroneDefense%
20Brochure.pdf
[30] DeDrone. (2020). RF-300 Data Sheet. Accessed: Jul. 28, 2020. [Online].
Available: https://assets.website-files.com/58fa92311759990d60953cd2/
5d1e14bc96a76a015d193225_dedrone-rf-300-data-sheet-en.pdf
[31] Rodhe and Schwarz. (2020). R&S Ardonis. Accessed: Jul. 28, 2020.
[Online].
Available:
https://scdn.rohde-schwarz.com/ur/pws/dl_
downloads/dl_common_library/dl_brochures_and_datasheets/pdf_
1/ARDRONIS_bro_en_5214-7035-12_v0600.pdf
[32] O. O. Medaiyese, A. Syed, and A. P. Lauf, ‘‘Machine learning framework for RF-based drone detection and identification system,’’ 2020,
arXiv:2003.02656. [Online]. Available: http://arxiv.org/abs/2003.02656
[33] M. S. Allahham, T. Khattab, and A. Mohamed, ‘‘Deep learning for RFbased drone detection and identification: A multi-channel 1-D convolutional neural networks approach,’’ in Proc. IEEE Int. Conf. Informat., IoT,
Enabling Technol. (ICIoT), Feb. 2020, pp. 112–117.
[34] Y. Liu, X. Wan, H. Tang, J. Yi, Y. Cheng, and X. Zhang, ‘‘Digital television based passive bistatic radar system for drone detection,’’ in Proc.
IEEE Radar Conf. (RadarConf), May 2017, pp. 1493–1497.
[35] J. Drozdowicz, M. Wielgo, P. Samczynski, K. Kulpa, J. Krzonkalla,
M. Mordzonek, M. Bryl, and Z. Jakielaszek, ‘‘35 GHz FMCW drone
detection system,’’ in Proc. 17th Int. Radar Symp. (IRS), May 2016,
pp. 1–4.
[36] Robin Radar Systems. (2020). Elvira. Accessed: Jul. 28, 2020. [Online].
Available: https://www.robinradar.com/elvira-anti-drone-system
[37] D.-H. Shin, D.-H. Jung, D.-C. Kim, J.-W. Ham, and S.-O. Park, ‘‘A distributed FMCW radar system based on fiber-optic links for small drone
detection,’’ IEEE Trans. Instrum. Meas., vol. 66, no. 2, pp. 340–347,
Feb. 2017.
[38] J. Colorado, M. Perez, I. Mondragon, D. Mendez, C. Parra, C. Devia,
J. Martinez-Moritz, and L. Neira, ‘‘An integrated aerial system for
landmine detection: SDR-based ground penetrating radar onboard an
autonomous drone,’’ Adv. Robot., vol. 31, no. 15, pp. 791–808, Aug. 2017.
[39] G. Fang, J. Yi, X. Wan, Y. Liu, and H. Ke, ‘‘Experimental research of
multistatic passive radar with a single antenna for drone detection,’’ IEEE
Access, vol. 6, pp. 33542–33551, 2018.
[40] M. P. Jarabo-Amores, D. Mata-Moya, P. J. Gómez-del Hoyo,
J. Bárcena-Humanes,
J. Rosado-Sanz,
N.
Rey-Maestre,
and
M. Rosa-Zurera, ‘‘Drone detection feasibility with passive radars,’’
in Proc. 15th Eur. Radar Conf. (EuRAD), Sep. 2018, pp. 313–316.
42656
[41] A. Crivellaro, M. Rad, Y. Verdie, K. M. Yi, P. Fua, and V. Lepetit,
‘‘A novel representation of parts for accurate 3D object detection and
tracking in monocular images,’’ in Proc. IEEE Int. Conf. Comput. Vis.
(ICCV), Dec. 2015, pp. 4391–4399.
[42] K. R. Sapkota, S. Roelofsen, A. Rozantsev, V. Lepetit, D. Gillet, P. Fua,
an…