Physical Layer Security in UAV Systems: Challenges and Opportunities
Xiaofang Sun, Derrick Wing Kwan Ng, Zhiguo Ding, Yanqing Xu, Zhangdui Zhong
11 Physical Layer Security in UAV Systems:Challenges and Opportunities
Xiaofang Sun,
Member, IEEE,
Derrick Wing Kwan Ng,
Senior Member, IEEE,
Zhiguo Ding,
Senior Member, IEEE,
Yanqing Xu,
Student Member, IEEE, and Zhangdui Zhong,
Senior Member, IEEE
Abstract
Unmanned aerial vehicle (UAV) wireless communications have experienced an upsurge of interestin both military and civilian applications, due to its high mobility, low cost, on-demand deployment, andinherent line-of-sight (LoS) air-to-ground channels. However, these benefits also make UAV wirelesscommunication systems vulnerable to malicious eavesdropping attacks. In this article, we aim to examinethe physical layer security issues in UAV systems. In particular, passive and active eavesdroppings aretwo primary attacks in UAV systems. We provide an overview on emerging techniques, such as trajectorydesign, resource allocation, and cooperative UAVs, to fight against both types of eavesdroppings inUAV wireless communication systems. Moreover, the applications of non-orthogonal multiple access,multiple-input and multiple-output, and millimeter wave in UAV systems are also proposed to improvethe system spectral efficiency and to guarantee security simultaneously. Finally, we discuss somepotential research directions and challenges in terms of physical layer security in UAV systems.
I. I
NTRODUCTION
In the past few years, heterogeneous data applications have emerged in wireless communica-tions, such as video streaming, e-health monitoring, video conferencing, etc. However, traditional
X. Sun, Y. Xu, and Z. Zhong are with the State Key Lab of Rail Traffic Control and Safety and the Beijing EngineeringResearch Center of High-speed Railway Broadband Mobile Communications, Beijing Jiaotong University, Beijing 100044, China(emails: { xiaofangsun, yanqing xu, zhdzhong } @bjtu.edu.cn).D. W. K. Ng is with the School of Electrical Engineering and Telecommunications, University of New South Wales, Sydney,NSW 2052, Australia (Email: [email protected]).Z. Ding is with the School of Electrical and Electronic Engineering, The University of Manchester, Manchester, UK (Email:[email protected]). a r X i v : . [ c s . I T ] S e p communication networks with fixed infrastructures are unable to meet the increasingly stringentquality-of-service (QoS) requirements in the fifth-generation (5G) and beyond 5G networks. As aresult, the development of unmanned aerial vehicle (UAV) has created a fundamental paradigmshift in wireless communication systems to facilitate fast and highly flexible deployment ofcommunication infrastructures. In particular, by exploiting the high maneuverability of UAV,communication links can be established ubiquitously, especially in temporary hotspots, disasterareas, and complex terrains. Compared to traditional terrestrial wireless communications, UAVwireless communications have the following unique features [1]: • On-demand and swift deployment:
UAV enables fast establishment of temporal commu-nication infrastructures in emergency scenarios where legacy or fixed infrastructures aredestroyed or do not exist, such as disaster rescue, remote sensing, firefighting, and others.The deployment of UAV facilitates cost-efficient and uninterrupted communications. • High flexibility:
Due to the fully controllable three-dimensional (3D) mobility, UAV caneither stay quasi-stationarily or cruise continuously to a dedicated location, depending onthe requirements of wireless communication systems. The movement of UAVs provides anew degree of freedom to offer efficient communications. • High probability of line-of-sight (LoS) air-to-ground channel:
According to the fieldtrial results [2], the LoS component dominates the air-to-ground channels in many practicalscenarios, especially for rural areas or moderately height UAV altitude. This channel charac-teristic leads to the fact that the channel state information (CSI) can be directly determinedby the position of each node and used to facilitate the design of high-speed communicationsystems. • Limited resources:
It is noted that onboard batteries of UAVs have limited energy storage.Moreover, the sizes of UAVs are usually small for on-demand deployment, and the weightsof UAVs for communications are typically not exceeding kg for both safety and energyconsumption issues [1]. These limitations directly restrict the communication, computation,and cruising duration capabilities of UAV.Despite the promising gains brought by UAVs, the open nature of air-to-ground wirelesschannels makes secure information transfer a challenging issue. Specifically, on the one hand,information signals transmitted over wireless LoS channels are likely to be intercepted by someundesired receivers which leads to a risk of information leakage. On the other hand, wireless UAV transceivers are vulnerable to malicious jamming attacks. Hence, security plays an extremelyimportant role in UAV wireless communications. Unfortunately, traditional encryption techniquesrequire high computational complexity leading to a large amount of energy consumption [3]which may not be suitable to UAV systems. As an alternative, physical layer security is compu-tationally efficient and effective in safeguarding wireless communication networks via exploitingthe inherent randomness of wireless channels. As a result, various physical layer techniques havebeen proposed in the literature for guaranteeing communication security, e.g., [3]–[11].Since the channel characteristics determine the performance of physical layer security, the LoSchannels as well as the mobility and flexibility of UAVs bring both opportunities and challengesinto the physical layer security design in UAV systems. In particular, on the one hand, theLoS channel condition in UAV-based communication systems may increase the vulnerability toeavesdropping. On the other hand, the fully controllable mobility of UAVs can be exploited toenhance communication security via adjusting its trajectory. For example, if the locations ofeavesdroppers are known, the UAV can avoid flying close to them to reduce the potential ofinformation leakage for guaranteeing secure transmission. Hence, introducing UAV to wirelesscommunication systems is a double-edged sword which requires a careful system design andthorough research.Motivated and inspired by the unique challenges and opportunities brought by UAV, this paperaims to provide an important overview on physical layer security in UAV wireless communicationsystems, which sheds light on potential researches in this field. In particular, the potentialsecurity attacks specifically in UAV wireless communication systems are discussed. Moreover,corresponding solutions are provided in terms of emerging techniques by considering both theadvantages and limitations of UAV. To further enhance the physical layer security, the applicationsof advanced 5G technologies, e.g., NOMA, 3D beamforming, and mmWave, are investigated inUAV wireless communication systems. Finally, potential research challenges that have not beenaddressed in the literature and some future directions in terms of secure communications in UAVsystems are envisioned. II. S ECURITY A TTACKS IN
UAV S
YSTEMS
A typical physical layer security communication problem comprises a minimum of three nodes,i.e., a legitimate transmitter, a legitimate receiver, and a potential eavesdropper, which can bemodeled by a wiretap channel [3]. In general, secrecy capacity and secrecy outage probability
Legitimate receiver
Ground BS
Legitimate receiver Legitimate receiver
Cooperative linkInformation transmissionInformation leakageJamming link
Malicious jammerEavesdropperUncertain area
Trajectory multi-UAV cooperation Initial location
Final location
Active eavesdropping
Passive eavesdropping
Figure 1. Illustrations of security issues in UAV wireless communication systems. are the two fundamental metrics to evaluate the physical layer security performance. Typically,denote C M and C E as the Shannon capacities of the main channel and eavesdropping channel,respectively. Consequently, the secrecy capacity C S is defined by [3] C S = [ C M − C E ] + , (1)where [ x ] + (cid:44) max { x, } . Notably, perfectly secure communication between a legitimate pair ispossible, when the eavesdropper’s channel is a degraded version of the main channel.In UAV-based communication systems, a UAV can act as either a transmitter or a receiver.On the one hand, it is noted that when the UAV is a legitimate transmitter, the associatedair-to-ground LoS channels facilitate the signal reception at both legitimate receiver and theeavesdropper, which may increase the vulnerability to potential eavesdropping. On the otherhand, when the UAV acts as a legitimate receiver, the existence of LoS channels enhance thephysical layer security especially against passive eavesdropping. As such, to investigate thesecurity issues in the presence of eavesdroppers, we mainly focus on the former scenario wherethe UAV acts as a legitimate transmitter. The potential security attacks in UAV systems areillustrated in Figure 1 and will be described in the following sections. A. Passive Eavesdropping Scenarios
In this scenario, passive eavesdroppers intend to intercept some confidential data withoutdegrading the received signal quality at the legitimate receiver. Since the eavesdroppers work in a passive manner and only intend to intercept the confidential messages, they usually remainsilent and their position information is not easy to be obtained by the UAV, which makes thesystem vulnerable to eavesdropping. In the following, we discuss three cases with differentavailabilities of the eavesdroppers’ position information.
1) Full Position Information of Eavesdroppers:
An optical camera or a synthetic apertureradar equipped at the UAV can help to detect and track the positions of potential externaleavesdroppers. If the eavesdroppers stay stationary, obtaining the complete positions of theseexternal eavesdroppers are possible at the UAV. As such, the security issues are likely to beaddressed in such scenarios via proper resource allocation design and taking advantage of theflexibility of UAV. However, obtaining such precise position information requires a high hardwarecost. Besides, the required new equipments impose an extra load to UAV which may increaseits energy consumption.
2) Partial Position Information of Eavesdroppers:
When the eavesdroppers stay quasi-stationaryor move in certain areas, obtaining precise position information of eavesdroppers is challengingand expensive in terms of hardware cost and energy consumption. Then, with the aid of theequipped camera or radar on the UAV, partial position information of these eavesdroppers ispossible to be obtained at the UAV via sensing and tracking. In the literature, worst-case secrecycapacity and secrecy outage probability are the two common metrics to evaluate the performanceof UAV systems with partial position information of external eavesdroppers [3], [7], [8].
3) Absence of Position Information of Eavesdroppers:
When eavesdroppers hide themselvesphysically, it is difficult for the UAV to detect and track them. This raises a problem thatthe completely absence of position information makes the UAV system extremely vulnerable toeavesdropping. In such scenarios, perfectly secure communications cannot be always guaranteed.
B. Active Eavesdropping Scenarios
Compared to passive eavesdroppings, active ones are more dangerous. The reasons are twofold.On the one hand, the active eavesdropping intends to attack the main channel by degrading thechannel capacity. For instance, active eavesdroppers transmit jamming signals to the legitimatereceiver for degrading the capacity of the main channel, i.e., C M . On the other hand, the activeeavesdropping may also aim to improve the capacity of the eavesdropping channel. As such,eavesdropping attacks caused by an active eavesdropper is more harmful than that of passiveones. Specifically, the LoS channel characteristic can improve the capacity of the eavesdropping channel when the UAV is a legitimate transmitter. According to the duplexing mode, we discussthe following two types of jamming attacks in UAV wireless communication systems: • Full-duplex (FD) eavesdropper:
A FD active eavesdropper transmits jamming noise andintercepts confidential signals simultaneously and independently [12]. In such scenarios,the UAV would increase its transmit power to improve the quality of the jammed channelwhich not only facilitates the eavesdropping at the FD eavesdropper due to the LoS channelcharacteristics, but also consumes more energy to combat the active eavesdroppings. • Cooperative half-duplex (HD) eavesdroppers:
Multiple HD eavesdroppers can mimic a FDeavesdropper via cooperation. For instance, some eavesdroppers transmit jamming to thelegitimate receiver, while the others intercept the confidential signal.In both situations, when the CSI of the legitimate receiver is available at the active eavesdrop-pers, the UAV systems become severely vulnerable, as the active eavesdroppers can efficientlyinterfere the legitimate receiver.III. P
HYSICAL L AYER S ECURITY T ECHNIQUES FOR
UAV S
YSTEMS
To enhance the physical layer security, the unique properties of UAVs, e.g., high mobilityand flexibility in positioning, can be exploited. Moreover, by using advanced resource allocationtechniques, the secrecy performance can be further improved. In the following, we discuss thespecific effective physical layer security strategies against the corresponding eavesdroppers listedin Section II in air-to-ground channels, as shown in Table I.
A. Anti-Eavesdropping Techniques1) Joint Trajectory and Resource Allocation Design:
In UAV systems, there are variousresources needed to be allocated, such as transmit power, cruising speed, time slot, and frequencybandwidth. It is noted that resource allocation affects the signal strength received at not only thelegitimate receiver, but also the eavesdroppers. In the following, we first discuss how trajectorydesign can be exploited to improve the physical layer security in UAV systems. We then providean overview on the joint design of resource allocation and trajectory for physical layer securityprovisioning. a) Trajectory Design Approach:
In practice, the initial and final locations of the UAV areusually predetermined. Besides, the UAV is also constrained by the maximum cruising speedand duration. These define the maximum service area. As a result, UAV trajectory design has
TABLE IS
PECIFIC S ECURITY A TTACKS AND P OSSIBLE S OLUTIONS IN
UAV W
IRELESS C OMMUNICATION S YSTEMS
Security attacks Passive eavesdroppings FD/Cooperative HD active eavesdroppingsFull PI Partial PI Absence of PI Full PI Partial PI Absence of PITechniques Joint design (cid:88) (cid:88) (cid:88) (cid:88) (cid:88) (cid:88)
Robust design (cid:88) (cid:88)
Artificial noise (cid:88) (cid:88) (cid:88) (cid:88)
Multi-UAV CoMP (cid:88) (cid:88) (cid:88) (cid:88) (cid:88) (cid:88)
Technologies Multi-antenna (cid:88) (cid:88) (cid:88) (cid:88) (cid:88) (cid:88)
NOMA (cid:88) (cid:88) mmWave (cid:88) (cid:88) (cid:88) (cid:88) (cid:88) (cid:88)
PI: Position informationJoint design: Joint trajectory and resource allocation design become an important research topic for realizing efficient UAV-based communication systems[4]–[6]. In terms of communication security, the UAV can fly close to the legitimate groundnode and away from the eavesdropper if it is possible. The principle is to carefully design thetrajectory of UAV, such that the legitimate link can be enhanced while the eavesdropping linkbecomes weaken. b) Resource Allocation:
Specifically, a joint design of trajectory and resource allocation isa promising approach to further enhance physical layer security, e.g., [5], [6]. The basic principleof the joint design is that when a UAV has to fly close to the eavesdropper, the UAV can decreaseor shut down the transmission power to reduce the potential of information leakage. At the sametime, the UAV flies away from the eavesdropper with its full speed for saving more time slotsfor the future. In contrast, when the UAV flies close to the legitimate receiver, the UAV usuallyslows down and increases its transmit power for the confidential information transmission.We provide an example to show how UAV can be adopted as a mobile relay to guaranteesecure communications via joint trajectory and resource allocation design. In the consideredscenario, a legitimate transmitter intends to serve a legitimate receiver in the presence of a passiveeavesdropper with the aid of a UAV. The UAV is introduced as a mobile relay to complete thedata delivery and to enable physical layer security. The ultimate goal of the system design is tomaximize the spectral efficiency of the system, while guaranteeing secure communications andsubject to certain practical constraints as in [6]. The joint trajectory and resource allocation design -2 0 2 4 6-101-101
UAV trajectoryEavesdropper trajectory k m k m k m Figure 2. The optimized UAV trajectories for the eavesdropper moves anticlockwise (a) and clockwise (b) around the legitimatereceiver with radius 1 km. The achieved system rate (c) and the corresponding cruising speed (d) via joint trajectory and resourceallocation design to enhance the physical layer security in UAV-aided wireless communication scenarios. can be obtained by iteratively solving the corresponding approximated optimization problem.Figure 2 provides some interesting insights into the trajectory and transmission performanceduring the operation period. In this figure, the legitimate transmitter and its receiver locate at (0 , , and (5 km, , , respectively, in a 3D Cartesian coordinate system. The position ofthe eavesdropper is assumed to be available at the UAV for the optimal design. The maximumcruising speed of the UAV is km/h.Figure 2(a) and Figure 2(b) depict the optimized trajectories of a UAV for two interestingscenarios, where the eavesdropper dynamically moves anticlockwise and clockwise, respectively,around the legitimate receiver with radius km and velocity km/h. Variables s t , s r , and s e rep-resent locations of the legitimate transmitter, legitimate receiver, and eavesdropper, respectively. q and q N represent the predetermined initial and final locations of UAV, respectively. Fromthis figure, we find that the UAV first flies close to the transmitter for caching enough data.Then it approaches the legitimate receiver and avoids any possible trajectories leading closeto the eavesdropper. Figure 2(c) and Figure 2(d) depict the capacities and the correspondingcruising speed of the UAV, respectively, via the optimal joint design. In these two figures, theeavesdropper locates stationary at ( km, , . C U denotes the instantaneous achievable rateat the UAV. From Figure 2(c) and Figure 2(d), we find that when the UAV locates closer tothe eavesdropper than to the legitimate receiver, the UAV either caches data from the legitimate transmitter by slowing down and then hovering above it or keeps silent and moves away from theeavesdropper. Differently, when the UAV is closer to the legitimate receiver, the UAV transmitsconfidential signal with a positive secrecy rate and tries to hover above the desired receiveras long as possible. These indicate that joint trajectory and resource allocation design in UAVsystems is an effective way to enhance the physical layer security and improve spectral efficiency.
2) Robust Joint Design:
When only partial or statistical position information of eavesdroppersis available, robust joint design can be exploited to facilitate secure UAV communications byconsidering the worst case scenario. For example, the uncertain location area of the potentialeavesdropper can be modeled by a region with a center which is the exact location of theeavesdropper where the length of its radius is related to the amount of uncertainty. Then, therobust design for the worst-case communication security is to guarantee the QoS of the systemwhenever the eavesdropper is located within the region. In general, the UAV with optimizedtrajectory would fly as close as possible to the legitimate receiver for enhancing the capacityof the legitimate channel while cruises away from the uncertain area of the eavesdropper(s) asmuch as possible. Also, a higher transmit power from the UAV is adopted when the UAV isclose to the legitimate user to exploit the short distance communication between the transceivers.However, when the UAV has to approach the uncertain area of the eavesdropper(s), it decreasesor shuts down the transmit power and accelerates away from this area to reduce the potential ofinformation leakage.
3) Artificial Noise:
To address the security issues caused by passive eavesdroppers with theabsence of their position information, one effective approach is to transmit artificial noise to thenull space of the legitimate pair’s channels via performing cooperation among multiple UAVs andto transmit confidential data only when UAVs are close enough to the legitimate receiver via jointtrajectory and resource allocation design. To this end, confidential data and artificial noise areavailable among the cooperating UAVs. As such, the capacity of the confidential signal throughthe legitimate channel, i.e., C M in Eq. (1), is increased but that through the eavesdropping channel,i.e., C E , is reduced. However, since artificial noise consumes some transmit power, this wouldleave a smaller amount of system power for allocating to the confidential information signal.Therefore, power allocation is an important issue in UAV-based wireless communications forimproving the system secrecy capacity. However, finding the optimal power allocation for UAV-based typically means solving non-trivial NP-hard optimization problems which are generallyintractable. Hence, computationally efficient suboptimal solutions should be proposed. B. Anti-Jamming Techniques
To address the security issues caused by jamming attacks, cooperation among multiple UAVs[9] can be exploited to enhance the physical layer security, since when multiple UAVs areavailable, cf. Figure 1, the degrees of freedom for optimizing the system resources are increased.The cooperation approaches can be summarized into the following two cases.For instance, cooperative multi-point (CoMP) transmission technique can be employed atUAVs. In particular, these multiple UAVs can form a virtual antenna-array to enhance the receivedsignal strength at legitimate receivers while degrade that at the eavesdroppers. Furthermore, withmultiple UAVs in the system, one can optimize their trajectories and resource allocations such thatsome UAVs transmit confidential signal to the ground legitimate receivers while the others sendjamming signals to confuse the eavesdroppers. According to the availabilities of eavesdroppers’positions, UAV can adopt different transmission schemes against active eavesdroppings: • When UAVs have the complete position information of the eavesdroppers, multiple UAVscan be employed to facilitate a joint trajectory and resource allocation design via forminga virtual antenna-array. Consequently, the confidential signal transmission can be enhancedby focusing the information energy beams to the ground legitimate receiver while reducingthe possibility of information leakage. • When UAVs have only partial position or statistical information of the eavesdropper, robustjoint trajectory and resource allocation design can be introduced into multi-UAV networksto improve the secrecy rate of the system or to reduce the secrecy outage probability belowa certain level [7]. • When the position information of the active eavesdropper is completely absent at UAVs,multi-UAV applying cooperative jamming to the orthogonal space of the legitimate receiver’schannel and position adjustment are effective methods to degrade the received signals ateavesdroppers and enhance that at the legitimate receiver [8]. Moreover, multi-antennatechniques can be adopted to utilize spatial degrees of freedom, hence enhance the qualityof the received signal at the legitimate receiver, and reduce the potential information leakageto active eavesdroppers via precise beamformer design. Furthermore, one can exploit thehuge bandwidth of mmWave frequency band to avoid potential eavesdropping and takeadvantages of line-of-sight dominated channels in mmWave systems for realizing highlydirectional transmission. -2 -1 0 1 2050100150200250 Figure 3. (a): The rate region achieved by NOMA and OMA with different maximum cruising durations. (b): The accumulateddata for the internal eavesdropper versus its displacement with different accumulated confidential data rate targets.
Remark:
We note that the the aforementioned anti-jamming techniques can be also adoptedto reduce information leakage against passive eavesdroppings. Besides, the techniques used foraddressing security issues in air-to-ground channels can be also applied to ground-to-air oneswhere the UAV is a legitimate receiver. For instance, the flexibility of the UAV can be lever-aged against eavesdropping and jamming attacks via proper trajectory planning or cooperativereceiving. IV. A
DVANCED A PPROACHES FOR
UAV S
YSTEMS
To further enhance the system secrecy performance, some advanced techniques can also beincorporated into the UAV systems [11], [13]. In the sequel, we will discuss the potentialapplications of NOMA, beamforming, and mmWave techniques in UAV systems for improvingthe performance of physical layer security.
A. Enhancing Physical Layer Security by NOMA
NOMA is viewed as a promising technique to provide superior spectral efficiency by mul-tiplexing information signals at different power levels [13]. Hence, it is expected that NOMAcan bring additional rate and robustness to enhance the achievable rate in UAV physical layersecurity communications. Consider a scenario where a UAV acts as a relay to facilitate datadelivery to two receivers with different security clearance levels within a maximum cruisingduration T . The receiver with a lower security clearance level is a potential eavesdropper since it has a strong motivation in intercepting signals intended to a receiver with a highersecurity clearance. Then, when the eavesdropper suffers from a bad channel condition, NOMAis adopted to forward both confidential and public information simultaneously. Otherwise, UAVonly broadcasts the public information for security issues. The mode selection between NOMAand unicast is chosen based on the results of the proposed resource allocation optimization. Inparticular, for maximizing the spectral efficiency, one needs to jointly optimize the transmissionscheme, resource allocation, and UAV’s trajectory. However, the coupled optimization variablesgenerally result in non-convex optimization problems which are difficult to solve optimally. Asan alternative, an iterative suboptimal algorithm based on successive convex approximation canbe employed to facilitate a computationally efficient joint design [6].Under the aforementioned optimization framework, Figure 3(a) depicts the optimal rate regionachieved by NOMA and OMA, respectively, with different cruising durations. Figure 3(b) depictsthe maximum accumulated public data rate achieved at the eavesdropper versus the horizontaldisplacement of the eavesdropper from the legitimate receiver. From Figure 3, we find thatNOMA scheme always outperforms OMA in all the considered scenarios, which demonstratesthe spectral efficiency advantage brought by NOMA in striking a balance between public datarate and confidential data rate. Moreover, based on Figure 3(b), we find that the accumulatedrate drops rapidly when the eavesdropper is sufficiently close to the legitimate receiver. In fact,when the secrecy rate requirement cannot be satisfied, we set the accumulated public data rateas zero to account for the penalty of the failure in guaranteeing secure communications. B. Enhancing Physical Layer Security by Multi-antenna Technology
Multi-antenna technologies have been widely considered in wireless communications due tospatial degrees of freedom for achieving high spectral efficiency [13]. Recently, in order toconcurrently improve the received signal power at the legitimate receiver and degrade the signalstrength received at eavesdroppers, multi-antenna technologies have also been investigated fromthe physical layer security aspect to enhance the secrecy performance and robustness of thesystem. In speak of the UAV wireless communication systems, the applications of multi-antennatechnology can be realized from the following two approaches, cf. Figure 4.
1) Traditional Two-Dimensional (2D) Beamforming:
The traditional 2D beamforming is aneffective way to improve the physical layer security. For example, when the UAV has thecomplete knowledge of the CSI, the beamforming direction can be simply set pointing towards
2D Beamforming
3D Beamforming
Azimuth angleElevation angle
Azimuth angle
Legitimate receiverEavesdropper
Eavesdropper
Null space
Figure 4. Illustration of the multi-antenna beamforming technique for enhancing physical layer security. the orthogonal space of eavesdropper’s channel and hence the information can be safely conveyed.However, it is obvious that this approach may not achieve the best secrecy performance of thesystem as the degrees of freedom are not fully utilized to improve the received signal powersat the legitimate users. Therefore, there exists a trade-off between improving the received signalstrength of the legitimate users and degrading the signal qualities of the eavesdroppers.
2) Three-Dimensional (3D) Beamforming:
Compared to the traditional 2D beamforming, 3Dbeamforming can generate separated beams in the 3D space simultaneously to provide a betterservice coverage [14]. Thus the 3D beamforming yields a higher system throughput and cansupport more legitimate users than that of the 2D beamforming. Generally, the 3D beamformingtechnique is more suitable for scenarios where the users are distributed in 3D space withdifferent elevation angles to their transmitter. Due to the high altitude of UAVs, the legitimatereceivers and the potential eavesdroppers can be easily separated by their different altitudesand elevation angles to the UAVs. Furthermore, LoS channel characteristic in UAV systemsenables effective beamforming in both azimuth and elevation domains. Specifically, narrow andprecise beamformer can be created to improve the transmission efficiency with respect to thedesired legitimate receivers while reduces the possibilities of information leakage to the potentialeavesdroppers. C. Enhancing Physical Layer Security by mmWave mmWave communications have been widely studied in the literature [13], since they are able tosupport high data rate by utilizing the abundant frequency bands. One of the major challenges formmWave communications is that its performance depends on the availability of LoS channels.Thus, compared to the conventional terrestrial communications, the inherent LoS channels inUAV systems facilitate the use of mmWave for high-data rate communications.As demonstrated in [13], the mmWave channels in UAV systems are very sparse in the angulardomain, as there are generally not many scatters around UAVs in the sky. Hence, one can takeadvantages of the specific channel characteristics in mmWave UAV systems for realizing highlydirectional transmission and exploit the huge bandwidth of mmWave frequency spectrum toavoid potential eavesdropping. For instance, active eavesdropping attacks in UAV systems canbe addressed by adopting frequency hopping (FH) technique to hide confidential data over anultra-wide mmWave frequency spectrum. There are two main advantages of this technique. First,confidential data can be hided from the eavesdropper by frequently hopping to different carrierfrequencies without exploiting the CSI. Second, frequency diversity can be exploited in FH toimprove the robustness against jamming active attacks.V. O
PEN I SSUES AND C HALLENGES
UAV-based wireless communications is a growing research area and handling the associatedsecurity issues is the key to unlock its potential. Despite the fruitful research in this area, thereare variety of challenges to be tackled, as shown in Figure 5. In this section, we provide andlist several open issues and challenges for future works.
A. Practical UAV Channel Modeling and Position Acquisition
In the literature, air-to-ground channel is generally modeled by the LoS channel. However,this model may not be accurate for some scenarios, such as urban areas. Consequently, moreefforts should be devoted to realistic channel modelings and to be verified by practice field testmeasurements. Moreover, the mobility of UAV has significant impacts on the CSI acquisitionfor both the legitimate users and eavesdroppers. This brings challenges from the physical layersecurity perspective, as efficient CSI acquisition algorithms usually exploit the physical propertiesof the wireless channels. Therefore, it is necessary to design a pragmatic CSI acquisitionmechanism to detect and track positions of the legitimate receivers and eavesdroppers. Legitimate receiver Active eavesdropper
Pilot contamination attacks
Eavesdropper Legitimate receiver
Blocked
Practical UAV channel modeling
Legitimate receiver
Malicious UAVMalicious UAV attacks
Pilot contamination
EavesdropperLegitimate receiver
Limited onboard resources
Information transmissionInformation leakage
Jamming link
Figure 5. Illustration of security challenges in UAV wireless communications.
B. Physical Layer Security Against Pilot Contamination Attacks
In practice, efficient beamforming for secure UAV communications requires accurate CSIwhich can be obtained by exploiting the pilot signals. However, in some scenario, active eaves-droppers intentionally transmit deterministic pilot samples which are identical to that transmittedby the legitimate transmitter to deceive the UAV for facilitating eavesdroppings [15]. As a result,the UAV designs an inappropriate transmission strategy which benefits the signal reception atthe eavesdroppers. For example, the UAV may misjudge the actual network environment andfly close to the eavesdropper for confidential data delivery which increases the possibility ofinformation leakage. Hence, investigating some efficient policies to eliminate the impact of pilotcontamination attack is challenging but important for safeguarding UAV systems.
C. Physical Layer Security Against Malicious UAV Attacks
The high mobility and flexibility of UAVs can be exploited not only to enhance physicallayer security, but also to intercept the confidential data and even perform jamming to reducethe quality of the legitimate link, which may bring more challenges than handling conventionalterrestrial eavesdroppers for safeguarding UAV systems. However, only limited researches havebeen devoted in this important aspect from communication theory perspective. Therefore, it isdesired to investigate advanced techniques in terms of the physical layer security to protectagainst malicious UAVs. D. Optimal Joint Trajectory and Resource Allocation Design
Compare to terrestrial systems, the UAV-based system brings a new dimension to enhancephysical layer security via designing the trajectory of a UAV. However, the designs of trajectoryand resource allocation are generally coupled together, which would make the design optimizationproblem intractable. Consequently, most of existing designs are suboptimal. Unfortunately, theperformance gaps between the optimal solution and existing suboptimal solutions are unclear.Therefore, to improve the performance of physical layer security in UAV systems for missioncritical applications, efficient algorithms are desired to strike a balance between computationalcomplexity and system performance.
E. Limited Onboard Resources
One critical obstacle in UAV wireless communication systems arises from the restricted flightduration, limited onboard energy and computational capability, etc, as the battery capacities,sizes, and weights of UAVs are all limited. As such, investigating advanced techniques toenable sustainable secure UAV communication are desired. For example, energy harvesting fromsolar and laser provides a viable solution to supply energy to UAVs on-the-fly. Besides, UAVcooperation can effectively leverage the onboard resources among all the cooperative UAVs toenable secure communication. VI. C
ONCLUSION
This article provides an overview and comprehensive discussions on the specific security issuesin UAV systems. Challenges and opportunities brought by UAVs for safeguarding UAV-basedcommunication systems via physical layer security are fully exploited. First, key problems interms of security attacks are revealed. Then, physical layer security approaches in UAV systemsare proposed to effectively prevent from both passive and active eavesdroppings. Particularly, weprovide an illustrative example on guaranteeing security provisioning by jointly designing UAV’strajectory and resource allocation, which demonstrates the advantages brought by the flexibilityof UAVs. Moreover, we propose to apply NOMA, MIMO, and mmWave techniques in UAVsystems to further enhance the physical layer security and the spectral efficiency. Finally, somepotential research directions and challenges are envisioned. VII. A
CKNOWLEDGMENT
This work was supported in part by funding from the UNSW Digital Grid Futures Institute,UNSW, Sydney, under a cross-disciplinary fund scheme, in part by the Australian ResearchCouncil’s Discovery Project (DP190101363), in part by the UK EPSRC under grant numberEP/P009719/2, and in part by H2020-MSCA-RISE-2015 under grant number 690750.R
EFERENCES [1] Y. Zeng, R. Zhang, and T. J. Lim, “Wireless communications with unmanned aerial vehicles: opportunities and challenges,”
IEEE Commun. Mag. , vol. 54, no. 5, pp. 36–42, May 2016.[2] 3GPP, “Technical specification group radio access network: Study on enhanced LTE support for aerial vehicles,” 3rdGeneration Partnership Project (3GPP), Technical Report (TR) 36.777, 2017, version 15.0.0.[3] A. Mukherjee, S. A. A. Fakoorian, J. Huang, and A. L. Swindlehurst, “Principles of physical layer security in multiuserwireless networks: A survey,”
IEEE Commun. Surveys Tuts. , vol. 16, no. 3, pp. 1550–1573, Third 2014.[4] Y. Cai, F. Cui, Q. Shi, M. Zhao, and G. Y. Li, “Dual-UAV enabled secure communications: Joint trajectory design anduser scheduling,”
IEEE J. Sel. Areas Commun. , vol. 36, no. 9, pp. 1972–1985, Sep. 2018.[5] A. Li, Q. Wu, and R. Zhang, “UAV-enabled cooperative jamming for improving secrecy of ground wiretap channel,”
IEEEWireless Commun. Lett. , vol. 8, no. 1, pp. 181–184, Feb. 2019.[6] X. Sun, C. Shen, T.-H. Chang, and Z. Zhong, “Joint resource allocation and trajectory design for UAV-aided wirelessphysical layer security,” in
Proc. IEEE Globecom workshops , Abu Dhabi, Dec. 2018.[7] J. Huang and A. L. Swindlehurst, “Robust secure transmission in MISO channels based on worst-case optimization,”
IEEETrans. Signal Process. , vol. 60, no. 4, pp. 1696–1707, Apr. 2012.[8] Y. Zhou, P. L. Yeoh, H. Chen, Y. Li, R. Schober, L. Zhuo, and B. Vucetic, “Improving physical layer security via a UAVfriendly jammer for unknown eavesdropper location,”
IEEE Trans. Veh. Technol. , vol. 67, no. 11, pp. 11 280–11 284, Nov.2018.[9] L. J. Rodriguez, N. H. Tran, T. Q. Duong, T. Le-Ngoc, M. Elkashlan, and S. Shetty, “Physical layer security in wirelesscooperative relay networks: state of the art and beyond,”
IEEE Commun. Mag. , vol. 53, no. 12, pp. 32–39, Dec. 2015.[10] Y. Wu, R. Schober, D. W. K. Ng, C. Xiao, and G. Caire, “Secure massive MIMO transmission with an active eavesdropper,”
IEEE Trans. Inf. Theory , vol. 62, no. 7, pp. 3880–3900, Jul. 2016.[11] N. Yang, L. Wang, G. Geraci, M. Elkashlan, J. Yuan, and M. D. Renzo, “Safeguarding 5G wireless communication networksusing physical layer security,”
IEEE Commun. Mag. , vol. 53, no. 4, pp. 20–27, Apr. 2015.[12] C. Liu, J. Lee, and T. Q. S. Quek, “Safeguarding UAV communications against full-duplex active eavesdropper,”
IEEETrans. Wireless Commun. , vol. 18, no. 6, pp. 2919–2931, Jun. 2019.[13] V. W. Wong, R. Schober, D. W. K. Ng, and L.-C. Wang,
Key Technologies for 5G Wireless Systems . Cambridge UniversityPress, 2017.[14] Y. Zeng, J. Lyu, and R. Zhang, “Cellular-connected UAV: Potential, challenges and promising technologies,”
IEEE WirelessCommun. Mag. , vol. 26, no. 1, pp. 120–127, Feb. 2019.[15] X. Zhou, B. Maham, and A. Hjorungnes, “Pilot contamination for active eavesdropping,”