1

A Survey of Optimization Approaches for Wireless

Physical Layer Security
Dong Wang, Bo Bai, Senior Member, IEEE, Wenbo Zhao, and Zhu Han, Fellow, IEEE

Abstract—Due to the malicious attacks in wireless networks, application of some emerging wireless networks and commu-
physical layer security has attracted increasing concerns from nication services. Therefore, the theories and technologies of
both academia and industry. The research on physical layer information security have attracted increasing concerns from
arXiv:1901.07955v1 [cs.IT] 23 Jan 2019

security mainly focuses either on the secrecy capacity/achievable

secrecy rate/capacity-equivocation region from the perspective both academia and industry recently.
of information theory, or on the security designs from the During last few decades, the information security mostly
viewpoints of optimization and signal processing. Because of its depends on the cryptographic encryption and decryption meth-
importance in security designs, the latter research direction is ods which are deployed at the upper layers of protocol stack.
surveyed in a comprehensive way in this paper. The survey The encryption-based security technologies have been shown
begins with typical wiretap channel models to cover common
scenarios and systems. The topics on physical-layer security to be effective in many cases, but their inherent vulnerabilities
designs are then summarized from resource allocation, beam- are heavy computation and key management costs which may
forming/precoding, and antenna/node selection and cooperation. result in high complexity and resource consumption [2]. As
Based on the aforementioned schemes, the performance metrics an alternative security technology, the physical layer security,
and fundamental optimization problems are discussed, which are based on the information theory framework, is to utilize
generally adopted in security designs. Thereafter, the state of the
art of optimization approaches on each research topic of physical the inherent randomness of the physical medium and the
layer security is reviewed from four categories of optimization difference between the legitimate channels and the wiretap
problems, such as secrecy rate maximization, secrecy outrage channels to guarantee secure information transmission [3].
probability minimization, power consumption minimization, and Compared with cryptographic approaches, as shown in Table
secure energy efficiency maximization. Furthermore, the impacts I, the physical layer security does not rely on the computing
of channel state information on optimization and design are
discussed. Finally, the survey concludes with the observations capability of the communication equipments, and thus has
on potential future directions and open challenges. the advantages of lower complexity and resource savings. It
has been shown from the viewpoint of information theory
Index Terms—Physical layer security, optimization, resource
allocation, beamforming, precoding, cooperative transmission that the physical layer security can achieve perfect secrecy
even if the eavesdropper has very strong computing capability.
Besides, the physical layer security has a performance metric
I. I NTRODUCTION for secrecy evaluation, i.e., equivocation rate which measures
With the rapid evolution of information and communication the uncertainty of the confidential message at eavesdroppers.
technologies, mobile Internet and Internet of things (IoT) have Furthermore, by exploiting the physical layer features, this se-
become indispensable in daily life. As the foundation of these curity technique can flexibly adjust transmission strategies and
networks, the cellular network has been designed to support parameters to accommodate the channel changes. In summary,
Internet connectivity and full interworking with heterogeneous physical layer security presents distinctive advantages and
wireless access networks [1]. This fact, therefore, leads to promising prospects. Therefore, the physical layer security can
complicated network architectures, network topologies, access be used as an effective supplementary for the cryptographic
technologies, service requirements, and mobile equipments techniques to further enhance information security.
while bringing serious security issues in wireless information The concept of secrecy communication was first proposed
transmission. How to guarantee the security of confidential in the pioneering work of Shannon in 1949 [4], in which
information has become the precondition to the commercial secrecy communication was investigated from the viewpoint of
information theory. It was proposed therein that the approach
The research is partially supported by National Natural Science Foun-
dation of China (NSFC) No. 61871401, Anhui Provincial Natural Sci- termed “one-time pad” could achieve the perfect secrecy.
ence Foundation No. 1708085MF139 and No. 1708085QF158, US AFOSR However, it was very difficult to apply this method in practice
MURI 18RT0073, NSF CNS-1717454, CNS-1731424, CNS-1702850, CNS- due to the intractable difficulties of key generation and man-
1646607, and ECCS-1547201.
D. Wang is with New Star Research Institute of Applied Technology, Hefei agement. Being different from the Shannon’s model of secrecy
230031, China (e-mail: [email protected]). communication, Wyner proposed the wiretap channel model
B. Bai is with Future Network Theory Lab, 2012 Labs, Huawei Technolo- in 1975 [5], in which the perfect secrecy could be achieved
gies Co. Ltd., Hong Kong, China (e-mail: [email protected]).
W. Zhao is with New Star Research Institute of Applied Technology, Hefei at the physical layer by utilizing the difference between the
230031, China (e-mail: [email protected]). legitimate channel and the illegitimate channel without any
Z. Han is with the University of Houston, Houston, TX 77004 USA (e-mail: key. In Wyner’s wiretap channel model, the signal received by
[email protected] <mailto: [email protected]>), and also with the Department
of Computer Science and Engineering, Kyung Hee University, Seoul, South the eavesdropper was a degraded version of the signal received
Korea. by the destination. The characteristic of signal degradation at
 2

TABLE I
C OMPARISON BETWEEN CRYPTOGRAPHIC ENCRYPTION AND PHYSICAL LAYER SECURITY

Cryptographic encryption Physical layer security

Theoretical basis Cryptography Information theory
Secrecy level Can be deciphered by brute-force Achieving perfect secrecy
computing
Computing ability requirements Heavily relying on the computing ability Being independent of computing ability
Encryption key management Heavy costs resulting from key generation, With no need of any key
management, and distribution
Evaluation criterion Being unable to accurately assess the Evaluating secrecy precisely by equivocation rate
leakage of confidential information
Adaptability to channel changes Poor channel adaptability Adjusting transmission strategies and parameters to
well adapt the channel changes

the eavesdropper made it possible to achieve secrecy at the metrics, optimization problems, and security solutions in this
physical layer. It was also proved by Wyner that the secrecy survey are intertwined with the secrecy rate/capacity which
capacity of a discrete memoryless channel was the maximum are based on information theory.
value of the difference between the mutual information of Many excellent surveys have been published in physical
the legitimate link and the mutual information of the wiretap layer security, which provide comprehensive overviews and
link. Thereafter, Csiszár and Körner generalized the degraded insightful comments to understand the fundamental principles,
wiretap channel to broadcast channel with confidential mes- technology status, and future trends in this field. In [13], the
sages, and analysed the secrecy capacity of a more general fundamentals and technologies of physical layer security are
(non-degraded) wiretap channel [6]. Following these works, reviewed comprehensively. Specifically, in [13], the technolo-
Leung-Yan-Cheong and Hellman investigated the Gaussian gies, challenges, and solutions are summarized from more
wiretap channel and derived the secrecy capacity which is the methodological viewpoints involving wiretap coding, multi-
difference of legitimate channel capacity and wiretap channel antenna and relay cooperation, physical-layer key generation,
capacity [7]. Nevertheless, the early research work cannot and physical-layer authentication. Moreover, we highlight the
be applied directly, since the physical layer security needed focused issues and the main contents of some published
suitable secure coding schemes to match the channel states. surveys in Table II. In contrast to existing surveys, our work
However, the secure coding technology was less developed tries to review the recent advances in physical layer security
in early stage, and the theories and technologies on physical from the perspective of system optimization and design. First,
layer security were thus believed to be impractical. Moreover, we summarize the research topics and the secure strategies
the fact that the encryption-based security technologies held that cover extensive problems in system optimization and
a dominant position for a long time affected the development design, such as secure resource allocation, signal processing
of physical layer security. In recent decades, the encryption- and cooperative diversity. Second, the performance metrics and
based security technologies have exposed some limitations the related optimization problem formulations are investigated
in practical applications. Meanwhile, the coding theories and to provide deep insights into secure transmission designs.
technologies have got a rapid development, which laid a solid Finally, we survey the state of the art of optimization and
foundation of physical layer security. Accordingly, more and design on each research topic of physical layer security from
more attentions have been paid on the physical layer security. four categories of basic optimization problems, i.e., maxi-
The studies on physical layer security can be roughly mization of achievable secrecy rate, minimization of secrecy
summarized from two main aspects: 1) the studies related outrage probability, minimization of power consumption, and
to secrecy rate/capacity from the perspective of information- maximization of secure energy efficiency (EE). In particular,
theoretic security, 2) and the studies related to system designs some optimization approaches and secure strategies which are
from the viewpoints of optimization and signal processing usually appeared in physical-layer transmission designs are
[8]–[12]. The first aspect mainly focuses on the secrecy summarized with detailed procedures.
capacity, achievable secrecy rate, and capacity-equivocation In summary, this survey provides a well-rounded overview
region based on the ideas of information theory. On the for newcomers to understand the optimization and design in
other hand, the second aspect mainly focuses on the secure physical layer security. The contributions of this survey is
strategy designs based on the techniques of optimization and based on the following work: 1) Summarizing general wiretap
signal processing. Because of the importance in practical channel models to cover the basic scenarios in this field,
security designs, our objective in this survey is to provide followed with usually appeared optimization approaches. 2)
a comprehensive overview on the optimization and design of Investigating hot topics in physical layer security from the
secure physical layer transmission. The investigations on the perspective of system optimization and design. 3) Seeking
topic that we just mentioned are based on the framework of deep insights into performance metrics to achieve different
information-theoretic security, since all involved performance requirements in system designs. 4) Reviewing the state of the
 3

TABLE II
B RIEF SUMMARIES ON EXISTING SURVEYS

Surveys Publications Focused issues Main contents

[1] IEEE Communications Security for long term evolution Security functionalities, security vulnerabilities, and existing
Surveys & Tutorials (LTE) and LTE-advanced Networks. security solutions.
[10] IEEE Communications Physical layer security in multiuser Security improvements in multi-antenna, broadcast,
Surveys & Tutorials wireless networks. multiple-access, interference, and relay channels, as well as
physical-layer key generation and secure coding.
[13] IEEE Communications Comprehensive overview on the Technologies, challenges, and solutions in physical layer security
Surveys & Tutorials fundamentals and technologies of are studied from the aspects of wiretap coding, multi-antenna and
physical layer security. relay cooperation, physical-layer key generation, and physical-layer
authentication.
[11] Proceedings of the IEEE Lessons learned from Designing secure wireless systems with unauthenticated entities by
information-theoretic security with cooperative jamming/relaying and interference alignment.
multiple wireless transmitters.
[14] Proceedings of the IEEE Security vulnerabilities, security Discussing the security requirements and attacks at each protocol
threats, and efficient defense layer, investigating the existing security protocols and algorithms,
mechanisms. while exploring the state of the art in physical layer security.
[15] Proceedings of the IEEE Physical layer security in the Surveying the advances and challenges in resource constrained
Internet of Things. secrecy coding and secret-key generation in the Internet of Things.
[8] IEEE Signal Processing Cooperative security at physical Guaranteeing information security by using cooperative techniques
Magazine layer. which consist of carefully designed coding and signaling schemes.
[9] IEEE Signal Processing Signal processing techniques for Enhancing physical layer security in multi-antenna systems by
Magazine secrecy in multi-antenna wireless beamforming/precoding with or without artificial noise.
systems.
[12] IEEE Communications Recent research on enhancing Signal design and optimization to increase secrecy based on
Magazine secrecy via cooperation. cooperative relaying and jamming.
[16] IEEE Communications A joint framework involving both Proposing a joint security scheme by exploiting the security
Magazine the physical layer and application capacity and signal processing technologies at the physical layer
layer security technologies. and the authentication and watermarking strategies at the
application layer.
[17] IEEE Communications Physical layer security for massive Discussing the passive eavesdropping and active attacks in massive
Magazine MIMO. MIMO systems while proposing three detection schemes to
identify the active attacks.
[18] IEEE Communications Physical layer security in cooperative Pure or hybrid relaying/jamming combinations for secrecy
Magazine relay networks. improvements with trusted/untrusted relays.
[19] IEEE Communications Physical layer security in the 5G The opportunities and challenges offered by the disruptive
Magazine network. technologies enabling 5G for achieving high physical layer security.
[20] IEEE Communications Challenges of physical layer security Identifying the important issues to apply physical layer security
Magazine in practical applications. into practice.
[21] IEEE Wireless Several prevalent methods to Classifying the methods of physical layer security into five major
Communications enhance physical layer security. categories while comparing their reliability, computational
complexity, and secrecy capacity.
[22] IEEE Network Diversity techniques to improve Exploiting MIMO diversity, multiuser diversity, and cooperative
physical layer security. diversity to secure wireless communications.

art of optimization and design in this field and the harmful optimization and design in physical layer security is reviewed
impacts of channel state information (CSI) on designing in Section V, followed with usually appeared optimization
security solutions. 5) Discussing future possible directions and approaches and security strategies. Section VI investigates the
open challenges. common assumptions of CSI and their negative impacts on
secure transmission designs. Future possible directions and
The remainder of this paper is organized as follows. In
open challenges are discussed In Section VII to provide some
Section II, typical wiretap channel models and optimization
lessons for newcomers. Finally, the survey is concluded in
concepts are introduced to cover common communication
Section VIII. A diagram is illustrated in Fig. 1 to show the
scenarios and optimization approaches. In Section III, the
outline and structure of this paper. In addition, abbreviations
research topics in physical layer security are investigated
used in this paper are defined in Table III.
from the perspective of secure resource allocation, beamform-
ing/procoding, and antenna/node selection and cooperation. In Notations: Throughout this paper, matrices and vectors are
Section IV, we seek deep insights into performance metrics denoted by bold uppercase letters and bold lowercase letters,
which can be adopted in all research topics to evaluate the respectively. x denote the set of optimization variable without
proposed secure transmission strategies. The state of the art of physical meaning. Mutual information, conjugate transpose,
 4

TABLE III A survey of optimization approaches for wireless physical layer security
A BBREVIATIONS AND THEIR DEFINITIONS
Section I: Introduction
Section II: Fundamentals
Abbreviation Definition
A. Wiretap channel models
5G The fifth generation
B. Concepts of optimization
AF Amplify-and-forward Section III: Research topics on system designs of physical layer security
AN Artificial noise A. Secure resource allocation
CSI Channel state information B. Secure beamforming and precoding

CI Channel inversion C. Antenna/node selection and cooperation

D. Joint strategies of several approaches
DC Difference of convex functions
Section IV: Performance metrics and basic optimization problems in physical layer security
DF Decode-and-forward A. Secrecy rate/capacity

EE Energy efficiency B. Secrecy outage probability/capacity

C. Power/energy consumption
GSVD Generalized singular value decomposition
D. Secure energy efficiency
I/Q In-phase and quadrature Section V: The state of the art of optimization and design in physical layer security

IoT Internet of things A. Secure resource allocation

LTE Long term evolution B. Secure beamforming and precoding

C. Antenna/node selection and cooperation
MIMO Multiple-input multiple-output
Section VI: The impacts of CSI on physical-layer security designs
MIMOME Multi-input multi-output multi-eavesdropper A. The perfect CSI of all channels

MISO Multiple-input single-output B. The imperfect CSI of wiretap channels

C. The unknown CSI of wiretap channels
MISOME Multi-input single-output multi-eavesdropper
Section VII: Discussions on future directions and challenges
MRC Maximal ratio combining A. The influences of wireless channels

MRT Maximum ratio transmission B. The impacts of adversary model

C. The influences of hardware impairments
mm-Wave Millimeter-wave
D. The joint designs of physical layer security and classic cryptographic security
OFDMA Orthogonal frequency division multiple access E. The global optimization with security, reliability, and throughput
QoS Quality of service F. The commercial application of physical layer security
Section VIII: Conclusions
RCI Regularized channel inversion
SISO Single-input single-output
Fig. 1. The structural diagram of this survey.
SIMO Single-input multiple-output
SDP Semidefinite programming
SDR Semidefinite relaxation broadcast wiretap channels, multiple-access wiretap channels,
interference wiretap channels, and relay wiretap channels [10],
SE Spectrum efficiency
etc.
SPCA Sequential parametric convex approximation
1) MIMO wiretap channels: The simplest network in phys-
SNR Signal-to-noise ratio ical layer security is composed of a transmitter, a legiti-
SINR Signal-to-interference-plus-noise ratio mate receiver, and an unauthorized receiver (eavesdropper),
S-DPC Secret dirty-paper coding in which confidential messages are exchanged between the
ZF Zero-forcing transmitter and the legitimate receiver while protecting from
the unauthorized receiver. In such a scenario, the terminals
may be equipped with multiple antennas. The typical chan-
and Euclidean norm are represented by I(·; ·), (·)H , and k · k, nel model for multi-antenna scenarios is the MIMO wiretap
respectively. The trace of a matrix is denoted by Tr(). W  0 channel which can cover the special models of single-input
means that W is a positive semidefinite matrix. single-output (SISO), single-input multiple-output (SIMO),
and multiple-input single-output (MISO) channels. In the
MIMO channel in which the transmitter, receiver, and eaves-
II. F UNDAMENTALS
dropper are deployed with nt , nd , and ne antennas, respec-
In this section, we give several typical wiretap channel mod- tively, the general expressions for the received signals at the
els to cover the common scenarios and systems considered in legitimate receiver and eavesdropper are, respectively, given
the survey, and introduce general concepts of optimization and by [10]
optimization problems to clarify the variables and parameters y d = H d x s + zd , (1)
in security designs.
ye = He xs + ze , (2)
A. Wiretap Channel Models where xs is the nt × 1 encoded signal with a covariance matrix
The typical wiretap channel models usually include constraint E{xs xHs } = Qx for Qx  0 or an average power
multiple-input multiple-output (MIMO) wiretap channels, constraint Tr{Qx } ≤ Pmax for a peak power Pmax . The
 5

nd × nt matrix Hd and the ne × nt matrix He are the channel the receiver and the eavesdropper, respectively. Some special
gain matrices to the legitimate receiver and the eavesdropper, cases of multiple-access channel in physical layer security
respectively. zd and ze are white Gaussian noise vectors at are also investigated, such as SISO multiple-access channel
the legitimate receiver and eavesdropper, respectively. This with an eavesdropper [25] and multiple-access channel with
wiretap channel model is typical, and to be widely investigated common and confidential messages [3].
in physical layer security. 4) Interference wiretap channels: The interference wiretap
2) Broadcast wiretap channels: The broadcast wiretap channel refers to the scenario where multiple links are simulta-
channels are raised in multi-user networks with more than two neously active in the same time and frequency slot, and hence
receivers where one transmitter delivers confidential informa- potentially interfere with each other [26]. At the same time,
tion to multiple users with the presence of multiple eaves- the communications over the multiple links are overheard by
droppers. We assume that there are one transmitter equipped an eavesdropper. We consider the interference wiretap channel
with nt antennas, I users each with ndi antennas, and J with K user pairs and an eavesdropper, where the source user
eavesdroppers each with nej antennas. In the downlink, the k, the destination user k, and the eavesdropper are deployed
transmitter transmits confidential messages to the legitimate with ntk , ndk , and ne antennas, respectively, k = 1, · · · , K.
users while preventing from overhearing of the eavesdroppers. The received signals of destination user k and the eavesdropper
This broadcast channel can be equivalent to a compound are, respectively, written as [27]
wiretap channel which is defined as [3], [23] K
X
ydi = Hdi xs + zdi , i = 1, 2, · · · , I, (3) ydk = Hdkk xsk + Hdkl xsl + zdk , (7)
l6=k
yej = Hej xs + zej , j = 1, 2, · · · , J, (4) K
X
where xs denotes the nt ×1 encoded signal for the confidential ye = Hel xsl + ze , (8)
l=1
messages which is subject to a covariance matrix constraint
E{xs xHs } = Qx for Qx  0 or an average power constraint where xsl is the ntl × 1 transmitted signal of source user
Tr{Qx } ≤ Pmax for a peak power Pmax . ydi and yej are the l with a covariance matrix constraint or an average power
received signals at user i and eavesdropper j, respectively. Hdi constraint. The ndk × ntl matrix Hdkl denotes the channel
is ndi ×nt channel matrix to user i and Hej is nej ×nt channel matrix from source user l to destination user k. The ne × ntl
matrix to eavesdropper j. zdi and zej are white Gaussian matrix Hel denotes the channel matrix from source user l to
noise vectors at user i and eavesdropper j, respectively. The the eavesdropper. zdk and ze are the white Gaussian noise
compound wiretap channel has several special cases including vectors at destination user k and the eavesdropper, respectively.
the parallel wiretap channel with two eavesdroppers, the A further model of interest is the interference channel with
fading wiretap channel with multiple eavesdroppers, and the separate confidential messages, in which each source message
wiretap channel with multiple receivers [3], etc. In addition, must be kept confidential from all other unintended users. A
another specific broadcast channel is the broadcast channel specific case of this channel model is studied in [3] where
with separate confidential messages of each user in which SISO interference channel is used to deliver two confidential
each downlink message must be kept secret from all other messages.
unintended users (each user is seen as an eavesdropper for 5) Relay wiretap channels: A typical cooperative wireless
messages not intended to it) [10]. network considering physical layer security is consist of a
3) Multiple-access wiretap channels: In the multiple-access source, a destination, a relay, and an eavesdropper, each with
wiretap channel, multiple transmitters transmit messages to nt , nd , nr , and ne antennas, respectively. The relay is operated
a legitimate receiver with the existence of an eavesdropper. in a decode-and-forward (DF) mode. In the first phase, the
There are K transmitters each with ntk antennas, one legit- source transmits the nt × 1 signal vector xs to the relay. The
imate receiver with nd antennas, and one eavesdropper with relay, the destination, and the eavesdropper receive the signal
ne antennas. Let us define nd × ntk matrix Hdk and ne × ntk as [28]
matrix Hek as the channel matrices from transmitter k to the yr = Hsr xs + zr , (9)
receiver and the eavesdropper, respectively. Then, the received (1)
signals at the receiver and the eavesdropper are, respectively, yd = Hsd xs + zd , (10)
expressed as [24] y(1)
e = Hse xs + ze , (11)
K
yd =
X
Hdk xsk + zd , (5) where the nr ×nt matrix Hsr , the nd ×nt matrix Hsd , and the
k=1
ne ×nt matrix Hse are the channel matrices from the source to
the relay, the destination, and the eavesdropper, respectively.
K
X zr , zd , and ze are the white Gaussian noise vectors at the relay,
ye = Hek xsk + ze , (6) the destination, and the eavesdropper, respectively. The relay
k=1 decodes the received signal and forwards it to the destination.
where xsk denotes the ntk × 1 encoded signal at transmitter Let the nd ×nr matrix Hrd and the ne ×nr matrix Hre denote
k with a covariance matrix constraint or an average power the channel matrices from the relay to the destination and the
constraint. zd and ze are the white Gaussian noise vectors at eavesdropper, respectively. In the second phase, the nr × 1
 6

transmitted signal vector xr of the relay is a new version of or approximated by convex problems. By solving the resulting
xs by using an encoding scheme. Then, the received signals at convex problems, we can get the optimal solution of the
the destination and the eavesdropper are, respectively, obtained original nonconvex problems. Moreover, to overcome the
as difficulties of solving nonconvex problems, some heuristic
(2)
yd = Hrd xr + zd , (12) algorithms can be designed based on convex optimization,
such as randomized algorithms in which an approximate
y(2)
e = Hre xr + ze . (13) solution to a nonconvex problem is found by drawing some
The other typical cooperative channel model is the amplify- number of candidates from a probability distribution, and
and-forward (AF) relay channel which is also investigated taking the best one found as the approximate solution [31]. In
extensively in physical layer security, such as in [29], [30]. addition, for nonconvex problems, the compromise is to give
up seeking the optimal solution. Instead, we seek a locally
optimal solution by combining convex optimization with a
B. Concepts of Optimization
local optimization method, where convex optimization can be
In this subsection, the concepts of optimization and opti- used for initialization of local optimization.
mization problems are introduced for understanding the survey 2) Optimization in physical layer security: Following the
easily. great progress in theories and algorithms of optimization, the
1) General optimization problem: A general mathematical system designs in physical layer security has greatly benefited
optimization problem can be formulated as [31] from recent advances to the point where optimization has now
min f (x) emerged as a major signal processing technique.
x  Towards general optimization problem (14) in physical
 h (x) ≤ b , i = 1, 2, · · · , m,
i i (14) layer security, the objective function f (x) may be the con-
s.t.
 g (x) = c , i = 1, 2, · · · , n, sidered performance metrics, such as secrecy rate/capacity,
j j
secrecy outage probability/capacity, power consumption, and
where x is the set of optimization variable. The function f (x) secure EE which will be elaborated in Section IV. The
is the objective function. The constraint conditions hi (x) ≤ bi optimization variable x may be the resources in the designs
and gj (x) = cj are the inequality and equality constraints, of secure resource allocation, beamformer/precoder in the
respectively. If there is no constraint, we say the problem is designs of secure beamforming/precoding, or candidates of
unconstrained. The optimization problem formulated in (14) antennas/cooperative nodes in the designs of antenna/node
describes the problem of finding an optimal x∗ that minimizes selection and cooperation. The secure resource allocation,
f (x) among all x satisfying the constraints hi (x) ≤ bi and beamforming/precoding, and antenna/node selection and co-
gj (x) = cj . Therefore, x∗ is called the optimal solution of operation mentioned here are the research topics in physical
the problem (14). layer security, which will be discussed in detail in Section III.
Convex optimization is an important class of optimization In physical layer security, the majority of optimization
problem. The standard convex optimization is defined as [31] problems are nonconvex due to the property of logarithmic
subtraction in secrecy rate/capacity. We can roughly list several
min f (x) optimization problems usually appeared in this field as follows.
x 
 h (x) ≤ b , i = 1, 2, · · · , m,
i i (15) • Integer programming in which some or all optimization
s.t.
 dT x = cj , j = 1, 2, · · · , n, variables are constrained to be integer values. This kind
j
of problems is usually raised in the designs of secure
where f (x) and hi (x) are convex functions. Comparing to subcarrier allocation and antenna/node selection.
problem (14), the convex problem has the characteristics that • Mixed integer programming that concerns the problems
the objective function and inequality constraint functions must having discrete and continuous variables. In joint subcar-
be convex while the equality constraint functions gj (x) = rier and power allocation, or joint antenna/node selection
dTj x − cj must be affine [31]. Convex optimization problems and beamforming, such problems are dealt with usually.
can be solved optimally by many efficient algorithms, such as • Difference of convex functions (DC) programming where
interior-point methods. If a practical problem can be formu- the objective function is a subtraction of two convex
lated as a convex optimization problem, the original problem functions. This feature fits with the definition of secrecy
can then be solved. Therefore, many problems can be solved rate/capacity. Therefore, DC programming is widely used
via convex optimization by transforming the original problem for solving the problems of secrecy rate maximization.
into a convex optimization problem. • Quadratic programming where the objective function has
Another class of optimization problem is nonconvex op- quadratic terms. This problem appears in the designs of
timization which covers the problems with nonconvex ob- secure power allocation and beamforming, such as the
jective function or/and nonconvex constraint functions. The typical optimization problem of power minimization.
nonconvex optimization problems are usually intractable. The • Semidefinite programming (SDP) which optimizes a lin-
complexity of global optimization methods for nonconvex ear function of the variables subject to linear equality con-
problems may grow exponentially with the problem sizes. straints and a nonnegativity constraint on the variables.
However, some nonconvex problems can be transformed into In physical layer security, some nonconvex problems are
 7

Secure strategies

Resource allocation Signal processing Antenna/node selection and cooperation

Power Subcarriers Antenna User Relay Jammer

Beamforming Precoding
allocation allocation selection selection selection selection

噯
Hybrid strategies: jointly using several approaches


Fig. 2. Secure strategies for improving physical layer security.

usually transformed into SDP to get an efficient algorithm

Power and energy  Frequency Resource block

that is easy to implement.
• Fractional programming which focuses on optimizing a
ratio of two nonlinear functions. The typical example
is EE maximization with the considerations of physical ĂĂ

layer security. Subcarrier Subcarrier Subcarrier

To cope with the nonconvexity of the optimization prob- Time

lems in physical layer security designs, many optimization

techniques have been proposed, such as dual decomposition,
alternating search, penalty function method, sequential para-
metric convex approximation (SPCA), semidefinite relaxation User User
(SDR), and so on, which will be discussed in Section V. Eavesdropper

III. R ESEARCH T OPICS ON S YSTEM D ESIGNS OF Fig. 3. An illustration of the multidimensional wireless resources in a multi-
P HYSICAL L AYER S ECURITY antenna multi-node OFDMA-based wireless network.

Many conventional physical layer technologies of wire-

less communications without secrecy consideration can be
Given the limited network resources such as bandwidth
redesigned for confidential information transmission under the
and energy, the main challenge of secure resource allocation
framework of physical layer security. From the perspective
is to utilize the limited resources as efficient as possible to
of system designs, the research topics on physical layer
achieve the requirements of some performance metrics, such as
security mainly focus on secure resource allocation, secure
secrecy rate, secrecy outage probability, power consumption,
beamforming/procoding, secure antenna/node selection1 and
and secure EE. Hence, many works have focused on the
cooperation, and the joint considerations based on the afore-
two basic problems of secure resource allocation that are the
mentioned strategies, as shown in Fig. 2.
subcarrier allocation and the power allocation in multicarrier
networks.
A. Secure Resource Allocation The subcarrier allocation aims at finding the optimal sub-
Resource allocation which has been widely used in the carrier usage policy that is able to effectively improve spectral
conventional communications without the consideration of efficiency and information security. Without loss of general-
secrecy [32], is also an effective way for enhancing physical ity, the secure subcarrier allocation is usually formulated as
layer security. The multidimensional wireless resources make a binary integer programming [33]–[36]. More specifically,
it possible to intentionally extend the difference between the whether or not a subcarrier is used for communication is
legitimate channel and the wiretap channel by secure resource specified by a decision variable α ∈ {0, 1}, with α = 1
allocation. The multidimensional wireless resources typically meaning that the subcarrier is used for transmitting and α = 0
contain the frequency, timeslot, and power in orthogonal otherwise.
frequency division multiple access (OFDMA) networks. In Adaptive power allocation among multiple carriers and
multi-antenna and multi-node wireless networks, the wireless nodes is another important method, which can be applied for
resources generally refer to the spatial degrees of freedom a further performance improvement [37]–[48]. Accordingly,
provided by multiple antennas and nodes, as shown in Fig. different strategies based on joint subcarrier and power al-
3. location have been proposed to achieve different design re-
1 Node selection usually adopted in multi-node scenarios includes user quirements in physical layer security [34]–[36], [49]–[51]. The
selection, relay selection, and jammer selection. joint subcarrier and power allocation are generally modeled
 8

low-complexity suboptimal algorithms have been proposed to

User simplify the beamforming designs [57], [58].
Relay
Source precoding Noteworthily, several existing beamforming techniques
precoding
which are simple but not optimal have been also adopted
widely in different scenarios of secure communications,
such as null-space beamforming (also named zero-forcing
A no

…
rti is

beamforming) [57]–[61] and maximum ratio transmission

fic e
ia
l

(MRT) beamforming [58], [62]–[64]. Null-space beamforming

Jammer
beamforming chooses the beamforming vector lying in the null space of
Eavesdropper
the eavesdropper’s channel vector. Then, the eavesdropper gets
Fig. 4. An illustration of secure beamforming and precoding in a multi-
nothing in the transmission process, such that the information
antenna and multi-node cooperative network. leakage is avoided. In optimization designs, nulling signal at
the eavesdropper can be expressed as a constraint, i.e.,

as mixed integer nonlinear optimization which is an NP- he wH = 0, (16)

hard problem in most situations. In practice, a number of where he and w denote the eavesdropper’s channel vector
optimization techniques have been proposed to provide simple and the null-space beamforming vector, respectively. MRT
and suboptimal solutions for such combinatorial optimization is another attractive beamforming scheme because of its low
problems [34], [35], [38], [49], [50]. computational complexity. MRT combined with maximal ratio
combining (MRC) can maximize the signal-to-noise ratio
(SNR) at the receiver and achieve a performance close to
B. Secure Beamforming and Precoding
channel capacity in low-SNR scenarios [65]. In particular,
Signal processing techniques, such as beamforming and the transmitter calculates its MRT beamforming vector, which
precoding which are popular in multi-antenna and multi-node only requires the knowledge of the channel from itself to the
cooperative networks, have been demonstrated as promising receiver. MRT beamforming can be expressed as
ways to achieve the physical layer security [9]. The de-
ployment of multi-antenna and multi-node cooperative net- hH
w= , (17)
works is thought to have great potential to enhance not only khk
transmission effectiveness and reliability but also wireless where w and h denote the MRT beamforming vector and the
security. It has been verified that collaborative beamforming legitimate channel vector from the transmitter to the intended
and precoding in multi-antenna and multi-node cooperative receiver, respectively.
networks can bring some benefits in terms of the secrecy Precoding is another important technology to achieve differ-
rate, secrecy outage probability, power/energy consumption, ent design objectives in physical layer security, which is espe-
and secure EE. cially appropriate for multi-stream data transmission or multi-
Beamforming and precoding technologies have been ex- user access. When the intended transceivers are equipped
ploited to achieve different performance requirements in se- with multiple antennas, the confidential messages of one or
cure transmission. Secure beamforming typically refers to multiple users can be spatially multiplexed onto multiple
one-rank transmission by which only single data stream is independent subchannels via precoding. By optimizing the
transmitted over multiple antennas or nodes, whereas secure precoder, the interested performance metrics of physical layer
precoding refers to multi-rank transmission by which more security can be achieved while the quality of service (QoS) can
than one data streams can be transmitted at the same time be guaranteed simultaneously [66]–[71]. The secret dirty-paper
[9]. Generally speaking, beamforming serves as a special coding (S-DPC) has been proposed to achieve the maximum
case of precoding. An illustration of secure beamforming secrecy rate in [72], [73]. However, the complexity of S-DPC
and precoding in multi-antenna and multi-node cooperative is computationally prohibitive, so that it is difficult to apply
networks is shown in Fig. 4, where the source precoding is this precoding scheme in practice. The complexity of a pre-
assisted by artificial noise (AN) and the intermediate nodes coding scheme may be crucial, which affects the application of
are used for relay precoding and jammer beamforming. precoding schemes in practice. In literature, due to the high
The main idea of secure beaforming is to compute the complexity of the optimal precoding in some scenarios, as
optimal beamforming vector for achieving some performance alternatives the suboptimal schemes have been developed to
metrics of physical layer security by enhancing the signal qual- reduce computational complexity and facilitate their practical
ity at the destination node and decreasing the signal quality at application [66], [67], [69], [70]. As a matter of fact, the linear
the eavesdropper. Most of the secure beamforming involves precoding techniques are also attractive alternatives because
solving optimization problems. Due to the special form of of their simplicity [66], [74]. As more simple linear precoding
logarithmic subtraction in the secrecy rate, the optimization techniques, generalized singular value decomposition (GSVD)
problems of secure beamforming are usually neither convex [26], [28], [74]–[77] and regularized channel inversion (RCI)
nor concave in many situations. Therefore, they can only be [78]–[82] have been extensively adopted in physical-layer
solved by numerical methods with high complexity, such as secure transmission. The GSVD is to simultaneously diago-
in [29], [52]–[56]. To mitigate the computational cost, some nalize the legitimate channels and the wiretap channels, such
 9

that a set of parallel independent subchannels is created to network, user selection determines which users should be
transmit the messages of different users [74], [83]. Channel scheduled for confidential data transmission. Typically, the
inversion (CI) precoding, sometimes known as zero-forcing user with the best channel quality is selected to improve
(ZF) precoding, is a popular and practical linear precoding secrecy rate or throughput [103]–[105]. This optimal selection
scheme which can control inter-user interference by canceling scheme depends on both the legitimate and wiretap channels.
all signals leaked to the unintended users. RCI based on CI Some suboptimal user selection schemes with considering
has better performance than plain CI by using a regularization wiretap links are also used due to their low complexity or
parameter. RCI can achieve a tradeoff among signal power, the unavailability of wiretap CSI [22], [103], [104], [106].
interference, and information leakage [78]. In addition, user selection/scheduling can also be used for
Using AN to deteriorate the quality of the received signals saving power with secrecy rate constraints or enabling the
at eavesdroppers is also a good way in physical layer security, largest possible user set with an effective transmission power
which is herein referred to as noise-assisted secure strategies. constraint [107]. In some situations, the legitimate channels
In such strategies, the transmitted signal is superimposed with to users may experience severe propagation loss and deep
AN [30], [42], [45], [62], [83]–[92]. This strategy is also fading, and such users may have little chance to be scheduled.
termed as “masked beamforming” in the multi-input single- Therefore, the fairness of user selection/scheduling needs to
output multi-eavesdropper (MISOME) wiretap channel [75], be considered. Two competing problems should be balanced
“masked precoding” in the multi-input multi-output multi- herein: achieving the optimal secure QoS while ensuring each
eavesdropper (MIMOME) wiretap channel in [76], and “AN user with certain opportunities to access networks [22], [108].
precoding” in [42]. In order to avoid interfering destination In multi-relay cooperative networks, the distributed relay
node, the simple but not optimal method is to let the AN lie nodes may provide spatial degrees of freedom which can be
in the null space of the signal space, i.e., satisfying exploited to improve secure QoS against the eavesdropping
attack. It is well-known that cooperative relaying with relay
hz = 0, (18) selection can bring some benefits in terms of rate, EE, and
where h and z denote legitimate channel vector and AN vector, security. More specifically, cooperative relaying combined
respectively [42], [87]–[89]. Furthermore, the AN can also be with relay selection has the potential of maximizing the se-
optimized globally to achieve the optimal secure performance crecy capacity [109], maximizing the Shannon capacity to the
[30], [62], [83]. destination node as well as minimizing that to the eavesdropper
[109], [110], reducing the secrecy outage probability [111],
[112], maximizing the SNR ratio of the destination node to
C. Antenna/Node Selection and Cooperation the eavesdropper [110], [113], or saving the limited power
In multi-antenna and multi-node wireless networks, antenna of network nodes [64], [114], [115]. Generally speaking, to
and node selection have been exploited to strengthen trans- strengthen the network security against the eavesdropping at-
mission reliability, which also have great potential to enhance tack, three relay selection schemes have been proposed, which
wireless security [22]. It has been verified that selecting are referred to as minimum selection considering only the
the proper antennas or nodes from the candidate set is a relay-eavesdropper links, conventional selection considering
simple but effective way to improve the performance of secure only the relay-destination links, and the optimal selection
transmission while saving resource. As a result, antenna/node taking the both links into account [110], [113]. In literature,
selection and cooperation have been considered and widely some heuristic algorithms have also been proposed for the
investigated in many works. optimal relay selection with different purposes.
MIMO technologies are believed to be one of the foremost Relay nodes can be used for not only cooperative relaying
technologies pertaining to physical layer security. In a MIMO but also cooperative jamming [10], [11], [116]. Cooperative
system, transmit antenna selection provides solutions to reduce jamming with jammer selection also has the ability to enhance
the hardware complexity resulted from large antenna arrays secrecy of wireless networks. This security-enhanced strategy
and radio frequency chains, insertion losses attributed to selects the jammers from trusted or untrusted intermediate
radio frequency switches, and feedback overhead needed for nodes to confuse eavesdroppers by transmitting artificial in-
transceiver communication [93]. In physical layer security, terference signals [117]–[120]. With regard to the untrusted
transmit antenna selection as a usual approach to exploit nodes which may be potential eavesdroppers, we should use
spatial degrees of freedom in multi-antenna scenarios, has been them discreetly. However, it has been verified in [121] that,
comprehensively investigated for maximizing the secrecy rate seeking for cooperation with the untrusted relay nodes can
[93], improving the SNR of the legitimate channels [94], [95], achieve a higher secrecy rate than just treating them as pure
and enhancing security from the viewpoint of secrecy outage eavesdroppers. In other words, the untrusted relays can also be
performance [96]–[100]. used for cooperative relaying while protecting the confidential
In multi-user networks, the randomness of users’ geograph- data from them [29], [52], [122]–[129]. Therefore, no matter
ical locations leading to random signal attenuation indepen- whether the relays are trusted or not, they can be used
dently across users, can also be used to enhance secure intelligently for cooperative relaying or jamming [129], [130].
performance [101]. Accordingly, user selection/scheduling as Moreover, cooperative jamming with the destination node can
a promising paradigm can be adopted to utilize the spatial also provide secrecy improvements, such as in [40], [127],
diversity in multi-user networks [22], [102]. In a multi-user [131].
 10

strategies have also been addressed for some specific sce-

ng e
yi iv
la at
narios to obtain secrecy improvements, such as cooperative

re per
o
Co
R beamforming and user selection in [106], [145], jamming-
aided beamforming/precoding in [63], [84], [85], [91], [146]–
R [148], joint power allocation and beamforming/precoding in
D [77], [149]–[151], etc.
S
Node clustering
Destination
Source
IV. P ERFORMANCE METRICS AND BASIC OPTIMIZATION
J E PROBLEMS IN PHYSICAL LAYER SECURITY
Co mmi

For the secure transmission designs, the choice of perfor-

ope ng
ja

Eavesdropper
rat

J mance metrics is remarkably critical. In physical layer security

ive

designs, there are several problems usually being raised from

different performance requirements:
Fig. 5. An illustration of joint relay and jammer selection and cooperation.
1) The transmission effectiveness of secure transmission
strategies, that is evaluated by the achievable secrecy
In practice, the joint relay and jammer selection is more rate/capacity.
effective for improving the secure performance of the whole 2) The reliability of secure transmission strategies, which is
network than using any single approach. As illustrated in Fig. measured by secrecy outage probability/capacity.
5, by using such a joint method, some proper intermediate 3) The power cost of secure transmission strategies, that is
nodes are selected to operate at a conventional relaying mode the minimum power consumption needed for ensuring the
for assisting the confidential data transmission between the secure QoS.
source node and the destination node. Meanwhile, another set 4) The EE of secure transmission strategies, which focuses
of intermediate nodes are selected as jammers to confuse the on the amount of secret bits transferred with unit energy
potential eavesdropper [132]–[138]. or the energy consumption required for sending one secret
bit.
To investigate these problems listed above, the correspond-
D. Joint Strategies of Several Approaches ing metrics termed as secrecy rate/capacity, secrecy outage
The secrecy improvements in physical layer can be sup- probability/capacity, power consumption, and secure EE, are
ported by secure resource allocation, signal processing, and usually adopted in system designs to evaluate the achievable
antenna/node selection and cooperation. Secure resource allo- performance of the proposed secure transmission strategies.
cation mainly focuses on resource usage policies by fully using More specifically, these performance metrics are usually taken
the multidimensional wireless resources involving frequency, as the optimization objectives for system designs in different
timeslot and power. Secure beamforming and precoding be- application scenarios.
longing to signal processing are to design beamformer and
precoder to well exploit the characteristics of multi-antenna
A. Secrecy Rate/Capacity
and multi-node settings which may form MIMO or virtual
MIMO networks. Antenna/node selection and cooperation aim Being similar to the data rate in conventional communica-
at selecting the proper antennas or nodes from the candidate set tions, the secrecy rate is a fundamental metric to assess the
to improve the performance of secure transmission. All of the transmission effectiveness of physical-layer secure strategies.
foregoing strategies can be carried out to strengthen informa- In physical layer security, the secrecy rate is defined as the
tion security while achieving the requirements of performance secret bits transmitted on the given channel per second, which
metrics and resource savings. In other words, based on the heavily depends on channel inputs. To evaluate the secrecy
ideas of these fundamental security strategies, we can design more conveniently and computation affordably, the Gaussian
some different transmission schemes to achieve the specific inputs as well as the achievable secrecy rate are usually
requirements of network performance subject to secrecy and adopted in many works [12]. The achievable secrecy rate can
resource constraints. be described as the difference between the achievable data
In order to achieve better performance of physical layer rate of the legitimate channel and the wiretap channel with
security, any single approach mentioned above might not be the Gaussian codebook, which is expressed as
sufficient. Therefore, the joint strategies based on some of the Rs = [Rm − Re ] ,
+
(19)
above approaches may be preferable in practical applications.
As in [36], [139], and [140], joint resource allocation and user where [x]+ , max{0, x}. Rm denotes the data rate of the
scheduling have been proposed to enhance physical layer se- legitimate channel from the source node to the destination
curity in OFDMA networks. Antenna selection combined with node. Re denotes the data rate of the wiretap channel from
beamforming/precoding has been demonstrated to be effective the source node to the eavesdropper. Clearly, the achievable
in secure MIMO system designs [93], [95], [141]. Distributed secrecy rate is a lower bound of the secrecy capacity [12]. In
beamforming with relay/jammer selection has been exploited practical designs, by some approaches as secure beamforming
in cooperative networks [142]–[144]. Additionally, other joint or resource allocation, a non-zero secrecy rate can be obtained
 11

since the wiretap channel is intentionally degraded while taken as the optimization objective in secure transmission
improving the quality of legitimate channel. designs with the consideration of channel fading [12].
Another metric closely related to secrecy rate is secrecy Towards secure communication system designs, the primary
capacity, which is defined as the upper bound of the secrecy concern is how much the secrecy rate can be achieved for
rate [3] [152]. More specifically, the secrecy capacity is the delivering the confidential data securely and reliably. This
maximum secrecy rate by which the confidential messages of problem can be modeled as maximization of the achievable
the source node can be securely and reliably transmitted to secrecy rate, that is to maximize the achievable secrecy rate
the destination node whereas the unauthenticated users cannot as much as possible by using some physical-layer technologies
obtain any useful information in this process. In Wyner’s such as resource allocation, beamforming/precoding, coopera-
pioneering work [5], the secrecy capacity of a degraded tive diversity, or other optimization algorithms. To maximize
wiretap channel has been given by the achievable secrecy rate, the most important factor is the
power limitation in addition to the bandwidth. Accordingly,
Cs = sup {I(X; Y ) − I(X; Z)}, (20) one common formulation of achievable secrecy rate maximiza-
p(X)
tion on the given channels generally aims at maximizing the
where X denotes the channel inputs at source node. Y and secrecy rate under the constraints of the maximum allowed
Z denote the channel outputs at the destination node and power. For instance, the achievable secrecy rate maximization
eavesdropper, respectively. I(·; ·) represents the mutual infor- in a relay network can be modeled as
mation. The secrecy capacity shown in (20) can be achieved 
(S) (j)

by choosing the optimal input probability distribution p(X). max Rs Pt , Pt
(S) (j)
Pt ,Pt ,j∈Ω
For any distribution p(X), the corresponding X, Y , and Z (S) P (j) sum
form a Markov chain [3] [152]. s.t. Pt + Pt ≤ Pmax ,
j∈Ω (23)
Based on Wyner’s results, Csiszár and Körner investigated 
 0 ≤ P (S) ≤ Pmax
(S)
,
a more general (non-degraded) wiretap channel and derived or
t
its secrecy capacity as [6]  0 ≤ P (j) ≤ Pmax
(j)
,j ∈ Ω,
t
max
where Psum denotes the maximum sum transmission power
Cs = sup {I(V ; Y ) − I(V ; Z)}, (21) (S) (j)
p(V,X) of all nodes in the relay network, and Pmax and Pmax denote
the maximum transmission power of the source and the jth
where V is an auxiliary input variable. By introducing appro- relay nodes, respectively. In existing literature, there are two
priate random variable V , the maximization in (21) can be kinds of power constraints in the problem of secrecy rate
implemented over all joint probability distributions p(V, X) maximization. One is the sum power constraints of all nodes
while forming a Markov chain V → X → (Y, Z). As to (S) P (j) sum
specified by the constraint Pt + Pt ≤ Pmax in (23),
the familiar Gaussian channel, the secrecy capacity has been j∈Ω
derived in [7] as following: and the other is the individual power constraint of each node
(S) (S) (j)
specified by the constraints 0 ≤ Pt ≤ Pmax and 0 ≤ Pt ≤
Cs = Cm − Ce , (22) (j)
Pmax , j ∈ Ω in (23). Noteworthily, beamforming may be more
where Cm and Ce denote the Shannon capacities of the effective for maximizing secrecy rate by strengthening signals
legitimate and wiretap channels, respectively. on a desired direction and suppressing/eliminating signals
The aforementioned secrecy rate and secrecy capacity have on undesired directions. When beamforming is considered
been investigated without considering the fading of wireless in such a relay network, the weight vector of all relays
channels. However, fading of wireless channels is an inevitable will be introduced to replace power, as investigated in [59]
issue in many situations, as stated in [152] in which three and [154]. The maximization of the achievable secrecy rate
standard fading models as well as the corresponding ergodic has been comprehensively investigated in many scenarios, for
secrecy rate/capacity have been well discussed, including the example multicarrier systems, multi-antenna systems, multi-
ergodic-fading model, block-fading model, and quasi-static node cooperative systems, etc.
fading model. When the channel fading is taken into consid-
eration, the average capability of secure communication over B. Secrecy Outage Probability/Capacity
fading channels should be evaluated, and the ergodic secrecy Due to the channel fading and imperfect CSI, secure trans-
rate or secrecy capacity is then a quite suitable metric for this mission may be broken. Therefore, it is of particular interest to
case [12], [153]. In practice, since achieving ergodic secrecy explore the secrecy outage behaviour of a secure transmission
capacity may be computationally difficult in many situations, strategy [153], [155]. Then, the secrecy outage probability is
the achievable ergodic secrecy rate is therefore adopted to an appropriate metric to characterize the probability that secure
measure the secrecy performance in fading scenarios. The transmission cannot be achieved. Precisely, the secrecy outage
achievable ergodic secrecy rate is defined as the difference probability is defined as the probability that a secrecy outage
between the ergodic rates of the legitimate and wiretap chan- event happens.
nels with Gaussian codebooks, which is more computationally There are two different definitions of secrecy outage events.
efficient in many cases. As the lower bound of ergodic secrecy The more popular one is that the secrecy outage happens
capacity, the achievable ergodic secrecy rate has usually been when the instantaneous secrecy capacity Cs drops below a
 12

target secrecy rate Rs0 , i.e., {Cs < Rs0 } [40], [155]–[159]. In Another important concept related to the secrecy outage
other words, the target secrecy rate is too high to be supported probability is the secrecy outage capacity Cout (ǫ), which is
by the current channel state, and the information security is defined as the largest secrecy rate that can be supported under
compromised. The secrecy outage probability of this definition a tolerable secrecy outage probability ǫ [12] [157] [162] [163].
is given by In other words, the secrecy outage capacity is the maximum
pout Rs0 = Pr Cs < Rs0 .


(24) achievable secrecy rate such that the secrecy outage probability
is less than ǫ, i.e.,
In (24), the outage events {Cs < Rs0 } happen whenever
the intended receiver does not receive the secret messages pout (Cout (ǫ)) = Pr {Cs < Cout (ǫ)} = ǫ. (27)
reliably (i.e., the message cannot be decoded correctly by
The practical significance of secrecy outage probabil-
intended receivers) or the message transmission is not perfectly
ity/capacity is that these definitions provide outage formu-
secure (i.e., some information may leak to eavesdroppers)
lations which give a more explicit measure of the security
[160], [161]. However, this definition does not distinguish
level. From the system design perspective, it is meaningful
between reliability and security. As a result, an outage based
to evaluate the secrecy outage behaviour of the proposed
on this definition does not necessarily imply a failure in
transmission scheme [161].
achieving perfect secrecy. To be specific, the outage events
For the optimization design in physical layer security, the
{Cm < Rs0 } mean that the secrecy rate cannot be supported
reliability of secure transmission which is generally measured
by the legitimate channels and the secure transmission would
by secrecy outage probability has also attracted increasing
be certainly suspended. Clearly, these suspension events fall
concerns. Ideally, the secure communication should be im-
within the outage events Cs < Rs0 due to Cm < Rs0

plemented without outage. Motivated by this observation, we
implying Cs < Rs0 , but it is clearly not a failure in achieving
expect to reduce the secrecy outage probability with the best
perfect secrecy [161]. Then, the outage probability in secure
effort. This raises the optimization problem of secrecy outage
transmissions can be more explicitly expressed as [33]
probability minimization subject to resource constraints. Tak-
0
, Rs0 = 1 − Pr Cm ≥ Rm 0
, Cs ≥ Rs0 ,


pout Rm (25) ing the relay network as an example, the minimization of the
0
secrecy outage probability can be roughly formulated as
where Rm is the target coding rate of the confidential mes-
0 0 0
sage, Rs ≤ Rm . The outage events {Cm < Rm } imply pout Rs0


min
0 (S) (j)
the legitimate channels cannot support the coding rate Rm . Pt ,P t ,j∈Ω
0 (28)
Consequently, at a target secrecy rate Rs and a target coding (S)
 0 ≤ P (S) ≤ Pmax ,
0 t
rate Rm , the reliable and secure transmission can only be s.t.
0  0 ≤ P (j) ≤ Pmax
(j)
, j ∈ Ω.
, Rs0 .


ensured at a probability 1 − pout Rm t
The other definition of secrecy outage is proposed in [161], In (28), the peak power of each transmission node is taken
which directly measures the probability that a transmitted into account to limit the excessive high power consumption
message fails to achieve perfect secrecy. In [161], consid- resulted from the improvement of the secrecy rate in mini-
ering the Wyner’s encoding scheme [5], the rate difference mizing secrecy outage probability.
Re , Rm − Rs is defined to reflect the cost of securing
message transmission against eavesdropping, where Rm and
Rs , respectively, denote two rates chosen by secure encoder, C. Power/Energy Consumption
namely, the rate of the transmitted codewords and the rate Power/energy consumption is a key consideration in
of the confidential information. The transmitted messages resource-limited scenarios such as battery-dependent net-
can be decoded correctly if Rm < Cm , whereas it fails to works. In general, the sustainability of secure communications
achieve perfect secrecy if Re < Ce . Hence, the secrecy outage in such networks is the most important concern. Therefore,
probability is defined in [161] as to reduce energy consumption and prolong network lifetime,
power/energy cost is one primary metric considered in physical
pout (Ce ) = Pr {Rm −Rs < Ce |message transmission} , (26) layer security designs.
which is a probability conditioned upon a message actually Before designing a secure transmission scheme with limited
being transmitted. If the source transmitter has no knowledge power and energy, we first analyse the factors of power
about the instantaneous CSI of the legitimate channel, the consumption in wireless networks [163]. According to [164],
transmission may always occur, so that the secrecy out- the total power consumption along the signal path can be
age probability in (26) then reduces to the unconditional divided into two main components: the power consumption
probability Pr {Rm − Rs < Ce }. More generally, when the of all the power amplifiers Pa and the power consumption
instantaneous CSI of legitimate channel is available, the source of all other circuit blocks Pc . The power consumption of
transmitter can decide whether or not to transmit with possibly all power amplifiers heavily depends on the output power of
variable rates according to channel conditions. Therefore, it is power amplifiers Pt , i.e.,
possible to reduce the secrecy outage probability by carefully Pa = Pt /η, (29)
designing the rate of the transmitted codewords Rm , the rate
of the confidential information Rs , and the condition for where η is the efficiency of power amplifier. The other circuit
transmission [161]. blocks include the basic circuits at the transmitter and receiver
 13

excluding power amplifiers, such as active filter, frequency schemes focusing on the minimization problem of power
synthesizer, mixer, intermediate frequency amplifier, analog- cost. In general, power minimization means consuming the
to-digital or digital-to-analog converter, and so on [163], [164]. minimum power to achieve the fundamental demand of secure
Accordingly, the power consumption of all other circuit blocks transmission such as the minimum target secrecy rate [53],
Pc can be roughly expressed as [164], [165] [59], [114], [154], the required SNR threshold of destination
node [91], [170], the given probability of secrecy [86], or other
Pc = Nt Pct + Nr Pcr + Pc0 , (30)
performance requirements. For example, in a relay network,
where Nt and Nr denote the numbers of transmitter antennas the basic formulation of power minimization can be expressed
and receiver antennas, respectively. Pct and Pcr denote the as
power consumed by the basic circuits at each transmit and
!
1 (S) P (j) (S) P (j)
receive chain, respectively. Pc0 denotes the power consumed min P = 2η Pt + Pt +Pc + Pc
(S) (j)
Pt ,Pt ,j∈Ω j∈Ω j∈Ω
by baseband circuits such as digital signal processing circuits.
It can be seen that Pct , Pcr , and Pc0 are independent of the s.t. Rs ≥ Rs0 .
secrecy rate. As a result, the total power consumption of a (33)
system can be given by The formulation in (33) is only a rough model, which can
P = Pa + Pc be specified in practical applications. For instance, when the
(31) beamforming is performed for minimizing the power con-
= Pt /η + Nt Pct + Nr Pcr + Pc0 .
sumption, the total power is then determined by the weights
The power consumption of a wireless communication sys- of beamformer [59], [154], [171].
tem can be usually formulated as (31). However, in a practical
scenario, there may be some variations in the power consump- D. Secure EE
tion model. For example, in a cooperative relay network, the
In the conventional communications without secrecy con-
power consumption model can be expressed as
  straints, the utilized efficiency of system energy referred to
1  (S) X (j)  X as EE is an important metric for green transmission strategy
P = Pt + Pt + Pc(S) + Pc(j) , (32) designs. When the security threats and energy limitations are
2η
j∈Ω j∈Ω considered jointly in wireless networks, it is significant to
where, Ω is the set composed of relay nodes and j denotes the design energy-efficient secure transmission strategies which
(S) (j)
jth relay node. Pt and Pt denote the transmission power should operate in a confidential and green manner. Therefore,
(S) from the perspective of green physical layer security, an ap-
of the source node and the jth relay node, respectively. Pc
(j) propriate metric for assessing the utilized efficiency of system
and Pc , respectively, denote the power of the basic circuit
energy is also of primary importance. In general, the utilized
blocks at the source node and the jth relay node, which can
efficiency of system energy can be measured by different met-
be obtained by (30). The factor 12 lies in the fact that the
rics from different viewpoints, such as the viewpoints from the
transmission is completed in two stages due to half duplex.
component level, equipment level, and system/network level.
The resource-limited regime motivates us to develop the
Towards the EE of system/network level, it aims at measuring
power-efficient transmission strategies which aim at minimiz-
both the energy consumed by all communication nodes and the
ing power consumption [114], [166], [167]. For this purpose,
performance experienced at the network level (i.e., capacity,
the power level of transmitters should be adjusted to save trans-
security, coverage, etc.). The EE of system/network level is
mission power while satisfying the target QoS requirements.
popular in transmission strategy designs.
It is worth noting that, although relay cooperation has the
There are two main metrics which have been defined for
potential of transmission effectiveness, reliability, and security,
evaluating the EE of novel techniques towards physical layer
relay nodes may consume additional power, such as the basic
security. One metric is the secure EE [64], [115], [172],
circuit power which is inherent in relay cooperation and
which is defined as the amount of secret bits transmitted with
unrelated with secrecy rate. Therefore, from the viewpoint of
unit energy consumption. Designing energy-efficient secure
transmission designs, the power adaptation and relay selection
transmission strategies with this metric, it is expected to
should be performed jointly to achieve the requirements of
maximize the secure EE. The resulting effect is that as much
power-efficient secure transmission.
confidential information as possible is transmitted with a given
It is noteworthy that a higher transmission rate of messages
amount of energy. Hence, given the amount of energy ∆E
can be achieved if no secrecy constraint is imposed. When
consumed in a duration ∆T , the secure EE can be defined as
secrecy is considered, the transmission rate of confidential
messages will decrease due to secure coding. Hence, higher Rs ∆T Rs
EB = = (bits/Joule) . (34)
power consumption is needed to ensure a higher level of ∆E P
secrecy at the physical layer [168]. The metric of secure EE is in fact the ratio of secrecy rate
For secure transmission designs in power-limited scenar- to total power consumption, which has been frequently used
ios, such as the transmission nodes powered by batteries in literature for investigating the EE of physical-layer secure
or energy harvesting devices [169], we should give priority communications [64], [71], [115], [172], [173]. This metric
to saving power and prolonging communication durations. is also termed as “secret bits per Joule”, since its unit is
These observations motivate us to design secure transmission bits/Joule.
 14

Another metric proposed to assess the EE of physical-layer achieving high secure EE but with too low secrecy rate. In
secure transmissions is the energy per secret bit, which is some literature [51], [178], Rs0 can be adjusted to balance the
suitable for evaluating the minimum energy required to send system performance between secure EE and secrecy rate.
one secret bit (i.e., minimum bit energy required for reliable It is pointed out that the aforementioned metrics and op-
communications under secrecy constraints). The precise for- timization problems are all based on information-theoretic
mula of this metric is the ratio of total power consumption to security, since those metrics and problems are intertwined with
secrecy rate [174] [175], i.e., the secrecy rate/capacity which is based on information theory.
P According to [179], another type of performance metrics for
EJ = (Joules/bit) . (35) secrecy is based on practical measures where the secrecy
Rs
level is quantified by the metrics that can be observed in
Noteworthily, these two metrics of secure EE are reciprocal practical communication scenarios, such as secrecy gap which
to each other. The resulting optimization problem by using one is usually characterized by bit error rate or packet error rate
metric is in general the dual problem of that by using the other versus SNR. To be specific, secrecy gap reflects the minimum
metric. Which metric is better in practice should fully consider required difference between the SNR of legitimate receiver
the practical scenarios, for reducing the difficulties of secure and eavesdropper for which secure communication is possible
transmission designs. As stated in [176], the metric of secret [179], [180]. This metric has also been used to make a
bits per Joule is more popular since it is convenient to capture quantitative measure for system designs, for instance in [180]–
the degree of proportionality between the energy consumption [182].
and different levels of load. This metric can reflect dynamic
network conditions considering energy consumption and se-
V. T HE STATE OF THE ART OF OPTIMIZATION AND DESIGN
crecy constraints in different situations of load. However, the
IN PHYSICAL LAYER SECURITY
metric of energy per secret bit is suitable to assess the network
EE only at a nonzero secrecy rate. In the previous section, we discussed the performance met-
In addition, it is obvious that the metrics of secure EE rics and fundamental optimization problems in physical layer
are closely related to the model of the power consumption. security. Each research topic of physical-layer security designs
The traditional energy-efficient technologies only consider the investigated in Section III involves extending these funda-
transmission power, but which is not the only part of power mental optimization problems to practical scenarios according
consumption in a networks. A holistic and system-wide power to specific application conditions and solving the resulting
model is imperative [177]. Therefore, the secure EE should be optimization problems to achieve the required performance
formulated with all power consumption including transmission metrics. In this section, the state of the art of optimization
power, basic circuit power, and signaling overhead in the entire and design in physical layer security will be summarized
network [177]. from the perspectives of the aforementioned research topics
In physical layer security, more power and energy, compared in physical-layer security designs. Each research topic will be
with the conventional communication without secrecy require- presented from four categories of fundamental optimization
ment, may be consumed to protect confidential information problems including maximization of achievable secrecy rate,
against eavesdropping. This observation can be verified by the minimization of secrecy outrage probability, minimization of
secrecy rate function shown in (19) where the information rate power consumption, and maximization of secure EE.
leaking to the eavesdropper generates extra consumption of
power and energy. This fact may increase the burden of power A. Secure Resource Allocation
and energy supplies, in particular in the scenarios with limited As a promising way for improving the performance require-
power and energy. When the limited power and energy become ments of physical layer security, secure resource allocation
the main factors for securing communications, the first con- has been extensively investigated for different purposes. As
cern, impelled by the requirements of “green communication”, discussed above, the designs of secure resource allocation
is to deliver confidential information with high secure EE as are usually performed by solving four optimization problems
much as possible. This motivation raises the maximization of which are related with the corresponding performance metrics.
the secure EE in physical layer security. Also taking the relay 1) Maximization of achievable secrecy rate: Many works
network as an instance, the mathematical formulation of secure focus on designing secure resource allocation strategies to
EE maximization can be roughly modeled as improve achievable secrecy rate. A conventional approach
max

EB = RPs
towards maximizing secrecy rate in multicarrier systems is
(S)
Pt ,P 
(j)
t ,j∈Ω
to globally allocate the limited power and subcarriers for all
(S) (S) transmission nodes. This goal usually leads to a mixed integral

 0 ≤ Pt ≤ Pmax , (36)

 programming in many scenarios, which has been investigated
(j) (j)
s.t. 0 ≤ Pt ≤ Pmax , j ∈ Ω, in many works [34], [35], [38], [49], [50], [183]. Such as


 R ≥ R0 .
 in [34], the resource allocation for a secure multicarrier AF
s s
relay communication system is investigated, in which decision
It is worth noting that the secure EE maximization should variables µsi ∈ {0, 1} and µri ∈ {0, 1} are defined for the
ensure the secure QoS requirement which is specified by source and the relay, respectively, for specifying the state of
the constraint Rs ≥ Rs0 in (36). Here, Rs0 is used to avoid communication on a carrier i. More specifically, if µsi = 1
 15

and µri = 1 then both the source and the relay transmit in
Original problem
respective slots, while if µsi = 1 and µri = 0 then only
the source transmits in its slot and it remains silent with the Lagrangian function
relay in the second slot. The case µsi = 0 and µri = 0
indicates no communication in both the slots and the case Dual problem
µsi = 0 and µri = 1 has no significance. Then, the resource
allocation strategy for maximizing secrecy capacity in such a Decomposition
relay-aided multicarrier system can be derived by solving the
typical mixed integral programming Master problem Updating the dual variables

P
max Ci (Psi , Pri , µsi , µri ) Information passing
Psi ,Pri ,µsi ,µri i
 P between the two layers
max
 µsi (Psi + µri Pri ) ≤ Psum Finding the optimal
(37)

 i
 Subproblem 1 … Subproblem N values of original
s.t. Psi ≥ 0, Pri ≥ 0 optimization variables




µsi ∈ {0, 1}, µri ∈ {0, 1}, Fig. 6. Dual decomposition approach for secure resource allocation.

where Psi , Pri , and Ci denote the source power, the relay
power, and the secrecy capacity on carrier i, respectively. ing secrecy outage probability. Considering a typical secure
Other specific formulations towards different scenarios have OFDMA downlink system in [33], the outage-optimal subcar-
also been explored in this areas. In [35], a secure resource rier allocation is addressed to minimize the secrecy outage
allocation policy is addressed for a downlink OFDMA-based probability pmout of each user m while guaranteeing that each
network with the coexistence of secure users and normal user has the identical probability to access each subcarrier n.
users which have no confidential messages and do not care The formulation of the problem is summarized as
about security issues. In [38], the transmission modes re-
ferred to as no communication, direct communication, and min {pm }∀m
out
relay communication are determined adaptively by subcarrier
P

 µmn ≤ 1
allocation while the optimal source and relay power allocation

 ∀m

(38)
s.t.
P
policy over all subcarriers is addressed to maximize the sum µmn ≤ 1
secrecy rate under a total power constraint. Jamming and AN- ∀n




aided resource allocation for sum secrecy rate maximization µmn ∈ {0, 1},

is, respectively, studied in [49] and [183], where the former
focuses on the OFDMA-based two-way relay wireless sensor where µmn are the decision variables with µmn = 1 meaning
networks while the latter focuses on the OFDMA systems with that subcarrier n P m. Otherwise, µmn = 0.
is assigned to user P
joint secrecy information and power transfer. For considering The constraints µmn ≤ 1 and µmn ≤ 1 imply that
∀m ∀n
the fairness of resource allocation in secure multiuser OFDMA each subcarrier can only be assigned to no more than one
downlink works, the work presented in [50] aims to assign user with identical probability. It is noted that, to deal with
subchannels and allocate power to optimize the max-min the difficulty of such a probabilistic integral programming, a
fairness criterion over the users’ secrecy rate. Besides, robust random bipartite graph approach is proposed with a logarithm-
secure resource allocation in relay-assisted cognitive radio polynomial complexity when applying parallel implementa-
networks is investigated in [184] considering the uncertainty tions. A more complicated formulation of probabilistic mixed
of CSI. integral programming is investigated in [186] to minimize the
To solve the problems of secure resource allocation men- secrecy outage probability of a wireless systems with adaptive
tioned above, the approach of dual decomposition is usually transmission rates and secrecy rates, in which a stochastic
adopted in many foregoing works. The basic idea of dual network optimization framework is introduced to overcome
decomposition can be summarized as: 1) constructing a La- the difficulty of such a problem.
grangian dual problem associated with the original problem The outage-optimal power allocation is also explored exten-
by transforming the constraints into the objective function sively. By deriving explicit expressions of the secrecy outage
in the form of a weighted sum, and 2) decomposing the probability, the closed-form solutions of the optimal power al-
Lagrangian dual problem into distributed subproblems which location are obtained to achieve high outage performance in an
are then coordinated with a high-level master problem by AF relay network with destination-assisted jamming [40], an
iterative alternating optimization between the two levels [185], AN-aided secure multi-antenna transmission coexisting with
as illustrated in Fig. 6. Based on dual decomposition, the randomly distributed eavesdroppers [45] [187], and a MISO
resource allocation in some secure scenarios can be solved system with a multi-antenna eavesdropper [90], respectively.
by different distributed algorithms which are efficient for The minimization of the secrecy outage probability is also
computing in many cases. raised in the scenarios of secure wireless information and
2) Minimization of secrecy outage probability: Secure re- power transfer in [188]–[190]. In [188], the transmission
source allocation is also an effective approach for minimiz- power allocation and power splitting ratio for AN signal are
 16

TABLE IV
T HE COMPARISON OF POWER - EFFICIENT RESOURCE ALLOCATION IN DIFFERENT SCENARIOS

Scenarios Wireless resources Assumptions of CSI Secure QoS constraints Solutions

MISO networks with The total power for Perfect CSI of legitimate chan- Received SNR at des- A closed-form solution
AN [86] secrecy information nel and unknown CSI of eaves- tination and a given
and AN dropper probability of secrecy
Multiuser MISO net- The total power for Perfect CSI of all channels Target secrecy rate Numerical analyses based on a line
works with jamming secrecy information search method
[192] and jamming signals
A non-orthogonal mul- Decoding order, Instantaneous channel gains of Target secrecy rate and A closed-form solution by problem
tiple access system with transmission rates, all users and the average chan- secrecy outage proba- simplification
multiple users [193] and power nel gain of the eavesdropper bility
Traffic offloading via Data rate and trans- Perfect CSI of legitimate chan- Traffic demand and se- Performing a series of equivalent
dual-connectivity in mission power nel and statistics CSI of eaves- crecy outage require- transformations and proposing an
cellular networks [194] dropper ment efficient algorithm to compute the
optimal offloading solution

jointly optimized to minimize the outage probability for delay- friendly jamming is considered, the power allocation strategy
limited secrecy information transmission based on the ap- is optimized to minimize the total power allocated to the
proaches of dual decomposition and alternating optimization. information signals and jamming signals while maintaining
In [189], the minimum secrecy outage probability is achieved secure QoS requirements. A non-orthogonal multiple access
by optimizing the optimal placement of energy harvesting system is considered in [193] where a closed-form solution
node with physical layer security considerations. In [190], is derived to minimize the transmission power. Additionally,
the secrecy outage probability minimization problem and the considering the application scenario where an user communi-
average harvested energy maximization problem in wireless cates simultaneously with a macro base station and a small-
information and power transfer systems are solved by an cell access point, a joint optimization of traffic scheduling
optimization framework of target secrecy rate and power and power allocation problem is formulated in [194] with
allocation ratio. the objective of minimizing the total power consumption
It is worth noting that the minimization of the secrecy while meeting both the user’s traffic demand and secrecy
outage probability is dual to the maximization of the secrecy requirement.
outage capacity which is another optimization design related 4) Maximization of secure EE: Secure resource allocation
to secrecy outage performance. As in [36], the packet data rate, is also effectively used for improving the EE of physical-layer
secrecy data rate, power, and subcarrier allocation policies of secure communications. To the best of our knowledge, the
an OFDMA DF relay network are designed to maximize the concept of secrecy capacity per unit cost is defined in [195]
average secrecy outage capacity by the dual decomposition to study the cost-efficient wide-band secrecy communications,
and gradient method. In [191], the solutions of the optimal in which the cost of the secrecy capacity may be the number
relay power allocations for a massive MIMO DF relay net- of channel use, the duration of transmission, or the amount
work are derived for maximizing the secrecy outage capacity of energy consumption. The research status of secure EE
and minimizing the interception probability, respectively. The maximization by resource allocation can be summarized from
results in [191] are expanded in [48], in which to cope with the the following aspects.
nonconvexity of the joint node power and transmission time • Multiuser multiple-access networks: The secure EE max-
allocation problem, the approach of alternating optimization is imization of an OFDMA downlink network is studied in
addressed by maximizing over some of the variables and then [51] where the power, secrecy data rate, and subcarrier
maximizing over the rest. allocation policies are optimized based on fractional
3) Minimization of power consumption: The power con- programming and dual decomposition. In a time-division
sumption of physical-layer secure communications can also multiple-access network considered in [196], the secure
be decreased by the designs of secure resource allocation EE measured by the average energy consumption of the
strategies. To be specific, by optimal resource allocation, we system per transmitted information bit is investigated by
can consume as less power as possible to achieve different using Markov decision process and cross layer design
requirements of secure QoS, as shown in Table IV. The use techniques, where information flow and joint optimization
of AN or jamming signals can deteriorate the wiretap channel, of higher and physical layer is permitted. To tackle the
but it also increases the total power consumption. Therefore, problem in [196], the strategies of packet scheduling
the optimal power allocation between the desired information and transmitter buffering are designed while the heuristic
and AN/jamming signals is very important for saving power. In algorithm of simulated annealing is used to solve the op-
a MISO system in [86], the optimal power allocation between timization problem due to its advantage to help avoiding
transmitted information and AN is developed for minimizing local minima.
the transmission power while ensuring a given probability • Multi-antenna networks: The energy-efficient resource
of secrecy. In [192] where a multiuser MISO network with allocation is carried out in multi-antenna networks in
 17

[197] with different CSI scenarios involving perfect CSI, is proposed to strike a balance between secure EE and
partial CSI, and statical CSI. The work is expanded in SE by maximizing the weighted product of secure EE
[198] by using the strategy of AN, while the fractional and SE.
programming and the sequential convex optimization tool It is observed that, the most of the secure EE maximization
are introduced to tackle the nonconvex problem. In [199], formulations are nonconvex, so that they are very intractable in
based on the optimization framework of [197] and [198], practice. Therefore, some nonconvex optimization methods are
two EE metrics are optimized, namely the metric of introduced to cope with the challenges, such as the fractional
secret bits per Joule and the metric of secret-key EE programming, penalty function method, alternating optimiza-
which is defined as the ratio between the system secret- tion, DC programming, etc. To be specific, the fractional
key capacity and the consumed power. In particular in programming can transform the secure EE function (which
[163], [173], [200], the optimization problems of energy- is a fractional function) into a parameterized polynomial
efficient secure communications are formulated by using subtractive form which can be tackled by the Dinkelbach
an specific secure EE metric which is therein defined as algorithm. The penalty function method is able to eliminate
the ratio of the secrecy outage capacity to the total power the nonconvex constraint of secrecy rate by incorporating
consumption. the constraint into the objective function. In some cases,
• Relay networks: In [64], [115], [172], the energy-efficient the optimization problem is nonconvex or unsolvable for all
power allocation is developed for DF, AF, and untrusted variables, but it is tractable when we tackle the problem
two-way relay networks, respectively. To deal with the with some of the variables and then tackle it with the rest.
nonconvexity of the problems, several optimization ap- Such characteristics are beneficial to implement alternating
proaches are jointly applied, which involve fractional optimization. Towards the optimization problem in which the
programming, penalty function method, alternating opti- objective function can be reformulated as a difference of two
mization, DC programming. The EE of repetition coding convex functions, the DC programming is an effective method
and parallel coding relaying under the partial secrecy which solves the problem iteratively by solving a series of
regime is investigated by power allocation in [201] based convex subproblems. The explanations of these optimization
on the fractional programming and a golden section methods can be found in Table V.
search algorithm.
• Cognitive radio networks: To implement the energy-
efficient secure communications in cognitive radio net- B. Secure Beamforming and Precoding
works in [202], the optimal power allocation and power The deployments of multiple antennas or nodes in wire-
splitting at the secondary transmitter are optimized un- less networks facilitate the technologies of MIMO or virtual
der secrecy constraints, while an EE Stackelberg game MIMO to be applied extensively, which provide abundant op-
between the primary and secondary transmitters is for- portunities to perform secure beamforming and precoding [8],
mulated for maximizing their utilities. In [203], the [9], [12], [22]. It has been demonstrated that, by beamforming
medium access probability and transmission power of and precoding in multi-antenna and multi-node cooperative
secondary transmitters are jointly optimized to maximize networks, we can obtain some benefits in terms of secrecy rate,
the secure EE of the secondary network. In [204], a secrecy outage probability, power consumption, and secure
secure EE maximization problem is established under EE. Naturally, to gain these benefits, the optimization designs
the constraints of data rate and transmission power of on beamforming and precoding can be solved with the four
the cognitive transmission as well as the interference performance metrics in practice.
limitation to the primary user, which is solved based on 1) Maximization of achievable secrecy rate: Following the
the fractional programming, penalty function method, and extensive applications of multi-antenna technologies, secure
DC programming. beamforming and precoding have been paid increasing con-
• The tradeoff between energy and secrecy: The tradeoff cerns for secrecy rate improvements [9]. It is verified in
between energy and secrecy also attracts many concerns [216] that exploiting space-time diversity at a multi-antenna
recently [175], [205]–[207]. In [175] and [205], the transmitter can enhance information security and information-
tradeoff between energy and secrecy is explored from hiding capabilities. After that, to improve the secrecy rate of
an information-theoretic perspective, while the metric of multi-antenna networks, some optimal or suboptimal schemes
partial secrecy is proposed to characterize the secrecy of secure beamforming/precoding have been proposed for
level of a communication system by looking jointly at multifarious scenarios based on different methods.
the application layer metric and physical layer secrecy a) Conventional beamforming/precoding: The conven-
metric. In [206], a framework is developed to study the tional beamforming/precoding schemes, such as MRT, sig-
spectrum efficiency (SE) and EE for secure transmission nal/AN null space, and GSVD, are applied separately or
in underlaid random cognitive radio networks, and the jointly for secrecy enhancements, due to the inherent simplic-
joint secure SE and EE optimization problem is formu- ity and easy implementation. For achieving a better secrecy
lated therein by using an unified secure SE-EE tradeoff performance, power allocation is usually optimized for these
metric. The energy-efficient secure communication in schemes. The MRT beamforming controls the beam towards
large-scale device-to-device underlaid cellular networks the intended user for strengthening its received signals. Since
is investigated in [207], in which a link adaptation scheme the MRT beamforming may lead to information leakage on
 18

TABLE V
T HE EXPLANATIONS OF SEVERAL OPTIMIZATION METHODS USED FOR SECURE EE MAXIMIZATION

Optimization
Problem formulations Problem transformations Algorithm procedures
methods
0)
1) For a given initial value x0 , calculate ε1 = h(x g(x0 )
; let
iterative index i = 1.
2) For εi , calculate the optimal solution xi by solving the
o Being related to the parameterized problem.
Fractional
n
h(x)
max f (x) = g(x)
parameterized problem 3) Stopping test with xi : If true, then stop; otherwise, go to
programming
max {h(x) − εg(x) : x ∈ D} step 4).
[208], [209] s.t. x ∈ D with parameter ε. i)
4) For obtained xi , calculate εi+1 = h(x
g(xi )
, i := i + 1, and
return to step 2).

1) Choose a small penalty factor τ0 and an increasing factor

Defining a penalty function ρ for updating τ . Let iterative index i = 0.
min f(x) L(x) , max{0, lk (x)}k
Penalty 2) For τi , calculate the optimal xi by solving the resulting
 l (x) ≤ 0 for the nonconvex constraints penalty problem.
function s.t.
k
lk (x) ≤ 0, and transforming 3) Stopping test with xi : If true, then stop; otherwise, go to
method  x ∈ D,
the problem formulation into step 4).
[210], [211]
min{f (x) + τ L(x) : x ∈ D}, 4) Update τ by τi+1 = ρτi , i := i + 1, and return to step 2).
where k = 1, · · · , m. where τ > 0 is a penalty factor.

1) Choose a starting point x0 = (y0 , z0 ) and let iterative

By partitioning the variables x into index i = 0.
two subsets y and z, the problem 2) For the given zi , find the optimal solution yi+1 of
Alternating can be iteratively solved by tack- min{f (y, zi ) : y ∈ D(y, zi )}.
y
min f (x) ling the following subproblems 3) For the given yi+1 , find the optimal solution zi+1 of
optimization
[212], [213] s.t. x ∈ D min{f (y, zi ) : y ∈ D(y, zi )} min{f (yi+1 , z) : z ∈ D(yi+1 , z)} .
z
y
4) Stopping test with (yi+1 , zi+1 ): If true, then stop; other-
min{f (yi+1 , z) : z ∈ D(yi+1 , z)} wise, let i := i + 1 and go to step 2).
z

1) Choose a starting point x0 and let iterative index i = 0.

min{f(x)=f1(x)−f2(x)} Being solved iteratively by tack- 2) For fixed xi , find the optimal solution xi+1 of
ling min{f1 (x) − f2 (xi ) − min{f1 (x) − f2 (xi ) − h∇f2 (xi ), x − xi i : x ∈ D}.
DC s.t. x ∈ D, h∇f2 (xi ), x − xi i : x ∈ D}, 3) Stopping test with xi+1 : If true, then stop; otherwise, go
programming
where ∇ denotes the gradient of to step 4).
[214], [215] where D, f1 (x), and a function and h·, ·i denotes dot 4) Let i := i + 1 and go to step 2).
f2 (x) are convex. product.

the direction to the eavesdropper, the AN null-space beam- b) Optimal beamforming/precoding: To achieve the
forming can then be exploited to disrupt the reception at the optimal secrecy performance, the strategy of beamform-
eavesdropper by emitting AN on the null space of legitimate ing/precoding is carefully designed by optimization ap-
channels. Such a joint scheme with MRT and AN null-space proaches. The precoding matrix design for maximizing the
beamforming is of particular interest in practice when the secrecy capacity Cs (W) in a standard three-node (two legit-
eavesdropper’s CSI is unavailable. If the transmitter has the imate users and an eavesdropper) MIMO wiretap network is
full CSI of the eavesdropper, the ZF beamforming can then formulated as [83]
be performed to overcome the faults of information leakage to
the eavesdropper by completely suppressing the beam towards max Cs (W)
W (39)
the eavesdropper. To tradeoff the intended received signal s.t. Tr(W) ≤ P max , W  0,
and information leakage to eavesdropper or other users, RCI
precoding is proposed based on a real regularization param- where W is the precoding matrix with the maximum power
eter which can be designed for secrecy rate improvements. constraint P max and the positive semidefinite constraint W 
When all nodes in a network are equipped with multiple 0. Such a nonconvex problem is solved by alternating opti-
antennas while the perfect CSI of all nodes is available, the mization and dual decomposition, while the resulting algo-
GSVD precoding can be implemented to decompose both the rithm is extended to the scenario with destination jamming.
legitimate channels and the wiretap channels into a set of In [68], the linear precoding strategies for secrecy rate maxi-
parallel independent subchannels which can be used separately mization in multiuser multiantenna networks are investigated
to transmit different messages. The works on the conven- in the broadcasting and multicasting scenarios, and an it-
tional beamforming/precoding schemes are compared in Table erative algorithm based on second-order cone programming
VI. Noteworthily, these conventional beamforming/precoding is proposed with low complexity and provable convergence.
schemes are suboptimal in many situations, and the optimal Focusing on the secure communications in dual-polarized
designs in this field have therefore attracted great interest. MIMO systems, a scheme of dual-structured precoding is
addressed in [217] in which a preprocessing matrix based on
the polarized array spatial correlation and a linear precoding
 19

TABLE VI
T HE COMPARISONS OF THE WORKS ON THE CONVENTIONAL BEAMFORMING / PRECODING SCHEMES

Schemes CSI conditions Expression/Constraints Explanations

H
Joint MRT and AN Legitimate CSI W = H /kHk and Controlling the beam towards the intended user while Emitting
null-space beamform- HZ = 0 AN on the null space of legitimate channels. The performance can
ing [42], [58], [62], be improved by power allocation between AN and information-
[63], [88] bearing signal.
ZF beamforming Legitimate and W = HH (HHH )−1 or Eliminating information leakage to the eavesdropper. This strategy
[58]–[61] wiretap CSI He W H = 0 is generally obtained by HH (HHH )−1 or optimized with the
constraint He WH = 0 in system designs.
RCI precoding [78]– Legitimate CSI W = HH (HHH + αI)−1 To balance the intended signal and information leakage by de-
[82] signing a regularization parameter α. The secrecy performance of
this strategy can be improved by power allocation.
GSVD precoding Legitimate and HW = UA and He W = Performing GSVD for matrix (H, He ), and returning the precod-
[26], [28], [74]–[77] wiretap CSI Ue Ae ing matrix W, unitary matrices U and Ue , nonnegative diagonal
matrices A and Ae . Power allocation can also be optimized for
secrecy improvements in this strategy.
*Notations: H, He , W, and Z denote the matrices of legitimate channels, wiretap channels, beamforming/precoding, and AN, respectively.

scheme based on different CSI are concatenated. The secure in the presence of a multi-antenna jammer [224]. To transmit
beamforming for typical three-node (two legitimate users and information securely in millimeter-wave (mm-Wave) MISO-
a relay) MIMO relay networks is explored in [52] and [29], OFDM systems with partial channel knowledge, a hybrid
where the untrusted relay is treated as an eavesdropper. To precoder is implemented in [225] by an iterative design with
reduce the difficulties of the joint designs in [52] and [29], the objective of minimizing the secrecy outage probability.
the alternating optimization is used to iteratively deal with the 3) Minimization of power consumption: In the existing
source and the relay beamforming in an alternate fashion. To literature, secure beamforming and precoding are also used
solve the resulting subproblems from alternating optimization, to support the designs of power minimization in different sce-
the SDP is introduced in the both works to transform a narios. The beamforming for minimizing transmission power
fractional quadratically constrained quadratic problem into in relay networks is investigated in [53], [59], [91], [154]
a SDP problem by the technique of SDR [218] and the with different constraints. The typical mathematical model for
rank-one matrix decomposition theorem [219]. Besides, the minimizing the total power of the source and relays under a
beamforming for maximizing the secrecy rate in simultaneous target secrecy rate constraint Rs ≥ Rs0 is given as [59], [154]
wireless information and power transfer is designed in [54],
[220], [221], where the optimal solutions are derived also
min{Ps + kwk2 }
based on SDR. More specifically, by relaxing the rank-one Ps ,w (40)
constraint, the considered optimization problems are therein s.t. Rs (Ps , w ) ≥ Rs0 ,
constructed as SDP problems which can be solved easily by
some existing optimization techniques and rank-one reduction where Ps and w are the source power and the relay weights,
[54], [220]–[222]. respectively. In particular in [91], the beamformer of the relays
2) Minimization of secrecy outage probability: In physi- is optimized to minimize the power allocated for transmitting
cal layer security, the potential of secure beamforming and confidential information, so that as much power as possible can
precoding for minimizing secrecy outage probability has also be used to transmit AN to confuse the eavesdropper. In [226]
been explored in recent years. Naturally, the existing beam- where a secure multiuser broadcast system is considered, the
forming/precoding schemes mentioned in the last subsection optimal precoding matrix at the base station and the jamming
can also be used to achieve the goal of secrecy outage proba- covariance matrix at the friendly jammer are jointly designed
bility reduction. As in [159], the AN-assisted beamforming is to minimize the total transmission power under the signal-to-
performed for degrading the eavesdroppers’ channels while the interference-plus-noise ratio (SINR) constraints at the users
optimal power allocation between the confidential information and eavesdroppers. In [227], the transmission beamforming is
and AN is obtained in closed form to minimize the secrecy rate performed for minimizing the power consumption of a full-
outage probability. In [156], the outage probability of secure duplex base station considering both self-interference mitiga-
transmission is minimized by the single-stream beamforming tion and physical layer security. Additionally, the physical
(based on MRT and ZF beamforming) and the use of AN layer security in satellite communication is considered in
in the null space of the legitimate channels. When only the [170] where the beamforming and power allocation under the
location information of the eavesdropper is available at the individual secrecy rate constraints are designed for minimizing
source user in [223], [224], the location-based beamforming is the overall transmission power used by all beams. In a new
optimally designed to minimize the secrecy outage probability cognitive radio network as described in [228], a cooperative
in Rician wiretap channels, while the resulting solution is beamforming scheme is proposed to minimize the transmission
extended to examine the solution of the optimal beamformer power of a secondary transmitter while providing different
 20

SINR for an eavesdropper, a primary receiver, and multiple precoder design in a conventional three-node (including a
secondary receivers. transmitter, a legitimate receiver, and an eavesdropper) MIMO
The problems of power minimization by beamform- wiretap channel is proposed based on the fractional program-
ing/precoding are also raised in simultaneous wireless infor- ming and Taylor series expansion. In [174], by providing a
mation and power transfer systems considering multifarious second-order approximation to the MIMO secrecy capacity
settings. with its first and second derivatives, the metric of minimum bit
• Multi-antenna broadcast networks: In such settings, si- energy is examined for secure and reliable communications in
multaneous wireless information and power transfer is the low-SNR regime while characterizing the tradeoff between
implemented by transmission beamforming which is de- EE and secrecy. A beamformer design is performed in [243]
signed to jointly or separately satisfy the constraints of se- for secure and energy-efficient wireless communication over
crecy rate, secrecy outage probability, energy-harvesting MIMO channels with multiple user pairs and an eavesdropper,
outage probability, and received SINR ratio [229]–[235]. where a path-following computational procedure is proposed
In order to achieve secure transmission, the transmission to cope with the intractable nonconvex problem and to yield
beamforming is also aided with AN strategy in many at least a locally optimal solution. In [244], the robust energy-
works [230]–[235]. efficient transmission design for MISOME wiretap channels is
• Distributed antenna systems: In [236], the beamforming investigated by the fractional programming and tight convex
and AN vectors are jointly optimized to minimize the relaxation, so that the primal fractional optimization problem
total transmission power while providing QoS for reliable is solved by solving a sequence of SDP problems. The energy-
communication and efficient power transfer in a given efficient beamforming for secure cognitive communication
time slot, in which the capacity-limited backhaul links is is raised in [245], in which the primal problem is tackled
taken into account. by the combined use of the fractional programming and
• Multi-cell multigroup multicast systems: In [237], two DC programming. In addition, in a MIMOME network with
different optimization targets are considered for a multi- simultaneous wireless information and power transfer [246],
cell multigroup MISO system, i.e., power minimization the transmission covariance matrices and power splitting ratio
and SINR balancing. The centralized and distributed for decoding information and harvesting energy are designed
beamforming algorithms are proposed for the considered jointly to maximize the secure EE, where the fractional
optimization problems, based on the techniques of SDR programming and alternating optimization are also employed
and alternating optimization. for handling the nonconvexity of the optimization problem.
• Cognitive radio networks: Simultaneous wireless infor-
mation and power transfer are raised in cognitive radio
C. Antenna/Node Selection and Cooperation
networks in [238], [239]. In [238], the total transmis-
sion power at the energy transmitter and the secondary Antenna/node selection and cooperation in multi-antenna
transmitter is minimized by a cooperative precoding and multi-node wireless networks have been well exploited
design while satisfying secrecy rate, energy harvesting, for achieving different performance requirements of physical
and interference temperature constraints. In [239], the layer security. Being similar to the former subsections, the
total transmission power of the secondary transmitter is state of the art of optimization designs in this research topic
minimized while ensuring that the QoS requirement on can also be reviewed from the four categories of optimization
secure communication is satisfied. problems.
It is noted that, in many works, the technique of SDR is ex- 1) Maximization of achievable secrecy rate: Great efforts
tensively adopted in the designs of transmission beamforming have been made for the optimization designs of antenna/node
[229], [231]–[235], [239], such that an approximation problem selection and cooperation to increase the achievable secrecy
can be directly obtained and solved by the method of SDP. In rate. Multi-antenna diversity can provide the gain of secrecy
general, the resulting relaxed problem by SDR cannot ensure rate by designing proper strategy of antenna selection, as
to get a rank-one solution. It always acts as an upper bound investigated in [93], [95], [247]. In multiuser scenarios, user
of the performance for the original problem [218]. In some selecting/scheduling can bring the improvement of secrecy
cases, the solution obtained by SDR is provably optimal, or rate by using multiuser diversity, such as the optimal and
the rank of the solution can be reduced by some techniques of suboptimal scheduling in a multiuser MISO system [145],
rank reduction. Noteworthily, solving the SDP problem may the maximum instantaneous SNR scheduling and approximate
result in relatively poor performance if SDP returns a high- proportional fair scheduling in a multiuser MISO system
rank solution. To overcome the difficulty, a method termed as with a multi-antenna eavesdropper [108], and the round-robin
SPCA [240], [241] is usually employed to find a suboptimal user scheduling as well as the optimal and suboptimal user
solution [229], [231]. The SPCA method approximates the scheduling in a cognitive radio network [103], [104].
nonconvex constraints by an upper convex estimate, and then In cooperative networks, the broadcast feature of wireless
results in a problem which can be solved directly. The two transmission results in two aspects, namely node cooperation
methods are briefly compared in Table VII. and data secrecy [248]. Node cooperation means that users can
4) Maximization of secure EE: The energy-efficient beam- help improve each other’s rate by intelligently using their re-
forming and precoding in physical layer security have also ceived signals. Data secrecy implies that the information leak-
been given many attentions. In [71], the energy-efficient age to the undesired users may cause some severe problems of
 21

TABLE VII
T HE COMPARISONS OF SDR AND SPCA

Optimization methods Problem formulations Problem transformations Comments

SDR [218], [242] By defining X = xxT which is equivalent to X The core idea of the method is that we
being a symmetric positive semidefinite matrix with drop the rank-one constraint to obtain
min xT A 0x rank one constraint rank(X) = 1, we get that a SDP problem. The SDP problem
x∈Rn
xT Ai x = Tr(Ai X). By ignoring rank(X) = 1, can be handled very conveniently by
s.t. xT A ix ≥ bi , we obtain a relaxed problem known as an SDP: readily available software packages.
where Ai are symmet- min Tr(A0 X) However, the resulting SDP problem
ric square matrices, i =
X  may lead to relatively poor perfor-
0, 1, 2, · · · .
 Tr(A X) ≥ b , i = 1, 2, · · ·
i i
mance if SDP returns a high-rank
s.t. solution.
 X  0.

SPCA [240], [241] By defining a function Li (x, ϕi ) which is a con- The basic idea of the method is that,
vex upper approximation of the nonconvex function at each iteration, we replace each of
min f (x) li (x), i.e., li (x) ≤ Li (x, ϕi ), the original prob- the nonconvex constraints by its upper
x∈Rn
lem can be approximated by the following convex convex approximation function with
s.t. li (x) ≤ 0, i = 1, 2, · · · problem: an appropriate ϕi . Thus, the method
min f (x) is required to iteratively solve a con-
where f (x) is convex, x∈Rn vex problem based on convex opti-
and li (x) is nonconvex. mization. The difficulty of the method
s.t. Li (x, ϕi ) ≤ 0, i = 1, 2, · · · ,
is to carefully choose the upper con-
where ϕi is a slack variable which is updated at each vex estimates and slack variables.
iteration.

information security. Accordingly, node cooperation and data to the destination as well as for minimizing that to the
secrecy have been studied jointly in multi-node cooperative eavesdroppers. According to [110] and [132], the relay
scenarios in recent years. The cooperative nodes act two roles selection schemes can be sorted into four categories, as
in physical layer security, including cooperative relaying and listed in Table VIII.
cooperative jamming [9]–[12], [18], [249]. Cooperative relay- • Cooperative jamming: When there is the direct channel
ing is to enhance the legitimate channels while cooperative from the source to the destination, the relays can be used
jamming is to degrade the wiretap channels. In practice, the as jammers to emit artificial interference, such that the
cooperative nodes may be trusted or untrusted. For the trusted channels to the eavesdropper are degraded and the confi-
nodes, they can be used for relaying and jamming separately dential information is protected against eavesdropping, as
or jointly [18]. As to untrusted nodes, seeking for cooperative depicted in Fig. 7(b). A simple but suboptimal jamming
relaying or jamming with them may be better than treating strategy is null-space cooperative jamming which emits
them as pure eavesdroppers [121], [122]. According to the artificial interference in the null space of the channels
roles of the cooperative nodes, there are generally four kinds of from the relays to the destination. Such a null-space
node-assisted transmission designs, which involve cooperative jamming strategy degrades only the wiretap channels
relaying, cooperative jamming, hybird cooperative relaying while with no influence to the legitimate channels. Trying
and jamming, and cooperative relaying with AN [12], [30], to obtain the optimal cooperative jamming designs, the
as illustrated in Fig. 7. solutions of jamming signal weights are elaborated in
[59], [147], and [154]. In [28], two types of cooper-
• Cooperative relaying: When the channels from the source ative jamming schemes referred to as full cooperative
to the destination are too poor or even nonexistent, signal jamming and partial cooperative jamming are proposed
retransmission by intermediate nodes is an effective way depending on that whether both the transmitter and the
for confidential data transmissions, as shown in Fig. 7(a). temporary helpers (which are acted by the source and the
Seeking for cooperative relaying with the intermediate destination) transmit jamming signals at the same time.
nodes, the confidential data can be delivered securely The secure transmissions with and without cooperative
and reliably, while some signal processing technologies jamming are compared in [39] based on the worst-case
can be applied into system designs to achieve both optimization. Using the intermediate nodes to relay or
the performance requirements and resource saving. The to jam, which is a better choice? Such a problem in-
typical cooperative relaying supported by beamforming to volving cooperative mode decision is discussed in [129],
improve secrecy rate is investigated in [29], [52], [58]– in which the performance comparison between the relay
[61], [154], [250], where the relays are trusted [58]– transmission scheme and the direct transmission scheme
[61], [154], [250], [251] or untrusted [29], [52]. The with jamming is examined in the distance normalized
optimal power control for multi-hop relaying is raised SNR regime. Additionally, in [129], it is also concluded
in [44]. The optimal relay selection and relay placement that, in the high distance normalized SNR regime, the
for cooperative relaying are concerned in [109] and [252], direct transmission scheme provides higher secrecy rate
respectively. In [110], three opportunistic relay selection compared with the relay transmission scheme, while
schemes are studied for maximizing the Shannon capacity
 22

Cooperative Cooperative
relaying jamming
Eavesdropper Eavesdropper
Source Source
Destination Destination

(a) Intermediate nodes used as relays for beamforming (b) Intermediate nodes used as jammers for beamforming

Cooperative
jamming
AN
Co Cooperative
op
rel erati relaying
ayi ve
ng Eavesdropper Eavesdropper
Source Source
Destination Destination

(c) Intermediate nodes grouped as relays and jammers for beam- (d) Intermediate nodes used for joint relay beamforming and AN
forming precoding

Fig. 7. Different strategies of intermediate node assisted transmission in cooperative networks.

in the low distance normalized SNR regime, the relay tial information and transmit AN simultaneously. To
transmission scheme outperforms the direct transmission be specific, the confidential information retransmitted at
scheme. each relay is superimposed with an AN. This strategy
• Hybrid cooperative relaying and jamming: A more subsumes all the above three designs and makes better
widely-used strategy is hybrid cooperative relaying and use of available degrees of freedom at relays [30]. As
jamming which is based on the combined application of a result, the combined designs of cooperative relay-
the two methods. In such a hybrid strategy, the inter- ing and AN precoding have been widely considered in
mediate nodes are grouped as relays and jammers. The physical layer security, such as in [30], [117], [253],
relays retransmit the received confidential information [254]. Noteworthily, it is generally challenging to jointly
to improve the signal quality at the destination while optimize AN precoding and relay beamforming [30].
the jammers continuously emit artificial interference to Therefore, some suboptimal but effective schemes have
confuse the eavesdroppers, as illustrated in Fig 7(c). This been proposed. The joint AN-aided beamforming and
may take the confidential information under protection power allocation are studied in [28], where a closed-form
in all stages of cooperative transmission. As in [132], jamming beamformer and a GSVD-based secure relaying
[133], [138], [142]–[144], the hybrid cooperative relaying scheme with each corresponding optimal power allocation
and jamming are investigated in different settings where are developed for the cases of single and multiple stream
joint relay and jammer selection are also taken into data transmissions, respectively. In [255], for a single-
account. A multiuser relaying scheme with the intended relay MIMO wiretap channel, an interference alignment
user jamming is proposed in [102] where the optimal approach is addressed to obtain a simplified and subop-
user selection is also considered in the sense of max- timal design of AN-assisted cooperative relaying.
imizing the secrecy rate. In [123], the untrusted two-
way relaying with friendly jammers is investigated based 2) Minimization of secrecy outage probability: The spatial
on Stackelberg game. In [146], a hybrid relaying and diversity provided by multiple antennas and nodes can be
jamming scheme with the optimal relay selection and effectively utilized for reducing the secrecy outage probabil-
power allocation is developed for maximizing the secrecy ity in multi-antenna and multi-node networks. The antenna
rate, in which the destination and the source are used as selection for enhancing the secrecy outage performance has
jammers to jam the eavesdropper in the first and second been investigated for MIMOME networks [100], [256], MIMO
phase of cooperative transmissions, respectively. relay networks [257], non-orthogonal multiple access systems
• Cooperative relaying with AN: In this strategy, as shown [258], and cognitive radio networks [259]. The antenna selec-
in Fig. 7(d), the relays are used to forward confiden- tion combined with AN is proposed for decreasing the secrecy
 23

TABLE VIII
R ELAY SELECTION SCHEMES

Schemes Mathematical formulations Achievable secrecy rates Explanations


Conventional selection k ∗ = arg max γk,D Cs = max{Ck,D } − Ck∗ ,E The relay which has the highest instantaneous SNR
k∈Ω k∈Ω
of the relay-destination links will be selected for
relaying.

Minimum selection k ∗ = arg min γk,E Cs = Ck∗ ,D − min{Ck,E } The relay that has the lowest instantaneous SNR
k∈Ω k∈Ω
of the relay-eavesdropper links will be selected for
relaying.
n 1+γ o
k,D
Optimal selection k ∗ = arg min 1+γk,E
Cs = Ck∗ ,D − Ck∗ ,E The optimal selection incorporates the quality of both
k∈Ω
links into the selection decision metric. The relay
that has the highest achievable secrecy rate will be
selected for relaying.
 
γk,D 
Suboptimal selection k ∗ = arg min E{γk,E }
Cs = Ck∗ ,D − E Ck∗ ,E The suboptimal selection scheme selects the appro-
k∈Ω
priate relay based on the statistical knowledge of the
relay-eavesdropper links. The scheme can avoid the
instantaneous estimate of the wiretap channels.
*Notations: Cs is the secrecy rate. k ∗ is the selected relay from the relay set Ω. γk,D and γk,E are the instantaneous SNR of the links from relay
k to the destination and the eavesdropper, respectively. Ck,D and Ck,E are the channel capacity of the links from relay k to the destination and
the eavesdropper, respectively. E{·} denotes expectation.

outage probability in secure two-way relaying communications requirements have been also studied in physical layer security.
in [260]. In multi-node scenarios, the relay and jammer As investigated in [114], the so-called power-efficient secure
selection in DF relay networks are studied separately or jointly communication is discussed with the objective of power mini-
for minimizing the secrecy outage probability in [112], [136]. mization by optimal relay selection. In [266], a secure adaptive
The best relay and user pair selection for minimizing the relay cooperation approach is developed to ensure wireless
secrecy outage probability of a multiuser AF relay network information security in an untrusted relay network with relay
are addressed in [113], in which the analytical expressions of energy harvesting, while a greedy battery-aware relay selection
the secrecy outage probability are also derived for the proposed scheme is proposed to minimize the power consumption in
three selection criteria. The selections of the transmission pro- such a network.
tocols are also investigated in literature [261], [262]. In [261],
4) Maximization of secure EE: It has been verified that
the secure transmission protocol which switches between DF
antenna/node selection and cooperation also can bring the
beamforming and cooperative jamming is proposed for the
gain of secure EE. In [267], the secure EE of a cooperative
purpose of maximizing secrecy rate and minimizing secrecy
MIMO relay network is investigated, in which transmit an-
outage probability in different communication scenarios. A
tenna selection and MRC are deployed at the transmitter and
secure opportunistic transmission protocol that aims at finding
the receivers, respectively. Considering three possible cooper-
an optimal scheme between direct transmission and relaying
ation scenarios in [268], namely the jammer only, relay only,
transmission, is developed in [262] to achieve the lowest
and the relay-jammer pair, the adaptive cooperation schemes
secrecy outage probability of cooperative wireless networks.
are addressed for energy-efficient physical layer security. In
Some works also consider designing the strategies of node
[269], hybrid full-/half-duplex receiver deployment strategies
selection and cooperation to reduce the secrecy outage proba-
are proposed for wireless ad hoc networks to optimize the
bility of cooperative energy harvesting networks [263]–[265].
network-wide secrecy throughput and network-wide secure
In [263], the relay and jammer selection are considered for
EE, respectively. The potential advantages of massive MIMO
the cooperative energy harvesting networks with a friendly
technologies are also explored for improving secure EE [270],
jammer. In [264], the secrecy outage probability of a wireless
[271]. In [270], the potential benefits of massive MIMO aided
powered communication network with an energy harvesting
heterogeneous cloud radio access networks are explored in
jammer is analyzed and minimized by optimizing the time
terms of the secrecy and EE. In [271], the advantages of
allocation between the two phases of information transfer
massive MIMO relaying are utilized to improve the secure
and energy transfer. The work presented in [264] is extended
EE which is specially defined as the ratio of the secrecy
to a more general multiuser situation with an additional
outage capacity to the total power consumption. Moreover,
consideration of jamming power allocation in [265].
the energy-efficient secure communication over a large-scale
3) Minimization of power consumption: Although multiple wireless network is studied by the combined application of
node cooperation can support the improvements of information game theory and stochastic geometry in [272]. An alternating
security, multiple nodes used for information transmission optimization scheme is proposed therein for maximizing the
may bring additional power consumption. In particular, some secure EE of the legitimate transmitters by controlling the node
cooperative nodes may consume high power but bring incon- activation probability, confidential message rate, redundancy
siderably improvement of secrecy. Accordingly, node selection rate, and the number of active antennas. In addition, an energy-
and cooperation for saving power while ensuring secure QoS efficient node activation game between the transmitters and the
 24

eavesdroppers is also studied therein, where the transmitters B. The Imperfect CSI of Wiretap Channels
and the eavesdroppers control their node activation probabil- In many situations, the perfect CSI of the main channel can
ities to maximize the secrecy EE and the eavesdropping EE, be easily obtained by channel estimation and CSI feedback,
respectively. whereas getting the perfect CSI of the wiretap channels is
very difficult or even impossible. In such cases, the imperfect
VI. T HE IMPACTS OF CSION PHYSICAL - LAYER SECURITY CSI of eavesdroppers may be obtained in practice, based on
DESIGNS the past channel observations or a priori knowledge of the
It has been discussed that the priori knowledge of the particular propagation environment [273], [274]. The uncer-
legitimate and wiretap channels’ CSI is very important for the tainties of the imperfect eavesdropper’s CSI can be generally
choices of secrecy metrics and the designs of secrecy strategies characterized by three ways. The first way is that the channel
[12]. To achieve the optimal performance of secure transmis- of eavesdropper follow some probability distributions [144],
sion, the perfect CSI of both the legitimate and wiretap chan- [251], such as the Gaussian distribution, Rayleigh distribution,
nels is indispensable for system designs. For getting the CSI of Rician distribution, and so on. In this way, only the statistical
the legitimate channels, some conventional methods (such as information of the eavesdroppers’ channels, i.e., the mean and
training/estimation and feedback), being similar to that in the covariance of the probability distribution, is available for the
traditional communications without secrecy constraints, can system designs, such as the assumptions in [40], [42], [51],
be used in physical layer security designs. However, due to [58], [64]. The second way to characterize the uncertainties
the existences of estimation error and feedback delay in some of eavesdroppers’ channels is termed as the deterministic
cases, it may be difficult in practice to get the perfect CSI of uncertainty model in some literature [30], [39], [58], [170],
legitimate channels. Regarding the CSI of wiretap channels, it [234], [253], [275]. In the deterministic uncertainty model
can be obtained perfectly when the eavesdroppers are also the which belongs to compound channel in information theory, the
legitimate users of the network but have different service from unknown wiretap channels are assumed to fall in a sphere or
that of the intended users. However, when the eavesdroppers a set. To be specific, the uncertainty region of eavesdropper’s
are passive, vicious or even hostile, it may be impossible to channels
√ is modeled as a sphere He with center h̄e and radius
get the perfect CSI of such eavesdroppers. According to the ǫ, that is [30], [39], [58], [170], [234], [253], [275]
above discussions, the following assumptions of CSI have been He = he |khe − h̄e k2 ≤ ǫ

considered in physical layer security, i.e., the perfect CSI of (41)
= h̄e + ve |kve k2 ≤ ǫ .

all channels, the imperfect CSI of wiretap channels, and the
unknown CSI of wiretap channels. In (41), he , h̄e , ve , and ǫ > 0 denote the real channel vector
of eavesdropper, the estimated channel vector of eavesdropper,
A. The Perfect CSI of All Channels the estimation error vector, and the channel mismatch, respec-
tively. By this model, we have that he ∈ He . The third way to
In the literature on physical layer security, the perfect CSI of
model the imperfect eavesdropper’s channels is based on the
all channels has been commonly assumed for designing the op-
imperfect channel estimate h̄e , the estimation error vector ve ,
timal transmission scheme which can match the instantaneous
and a scalar κ ∈ [0, 1] for indicating the degree of channel
changes of channel states, such as in [35], [38], [44], [52], [55],
knowledge. This model can be expressed as [156], [276]
[59]–[61], [109], [154]. In fact, the perfect CSI including that √ √
of eavesdroppers, can be obtained at all communication nodes he = κh̄e + 1 − κve . (42)
in some situations. For instance, the eavesdropper is active
In (42), if κ = 1, it means that the eavesdropper’s CSI is
in the network and its transmissions can be monitored. This
perfect, while if κ = 0, it implies that we fail to get any CSI
case arises particularly in the practical applications combining
of the eavesdroppers.
multicast and unicast transmissions, in which the user plays
In some worse cases, the perfect CSI of both legitimate and
double roles as legitimate receiver for some signals and
wiretap channels is unavailable due to limited feedback or
eavesdropper for others [59]. Alternatively, the eavesdropper
other reasons, such as discussed in [40], [45], [56]. Then, the
is also a legitimate user of the network whereas its service
uncertainties of legitimate channels can also be characterized
differs from that of the intended user [61]. In other words,
by the three methods mentioned above. It is worth noting that,
instead of eavesdroppers, there can be friendly nodes in the
towards the uncertainties of real channels, the robust secure
network that are not supposed to hear certain messages. This
designs are commonly performed to ensure achieving the
case arises often in military communications, where lower
security, reliability, and robustness of information transmission
level network users can only access to less information [154].
[30], [39], [56], [234], [253], [275].
Because the confidential information of the source user is
expected to be received only by the intended user, the other
users (they are even legitimate and friendly) in the network C. The Unknown CSI of Wiretap Channels
should be treated as eavesdroppers for secure transmission The assumption on the perfect CSI of all channels is
designs. However, such legitimate and friendly users can feed commonly used for calculating the instantaneous secrecy ca-
back the perfect CSI to transmitters. Accordingly, the optimal pacity and secrecy rate which are needed for instantaneous
secure transmission designs can be performed with the perfect optimization designs. Using the perfect CSI, the security and
CSI of all channels. reliability of information transmission can be guaranteed by
 25

secure coding and rate adaptation. However, a more practical and eavesdroppers, as well as the node mobility model, may
assumption is that the CSI of wiretap channels is completely have important impacts on wireless channels, which also need
absent due to the concealment and hostility of eavesdroppers to be considered in secure transmission designs. Besides, it is
[47], [91], [256], [277]. Moreover, whether there exists any already known that the secure strategy designs heavily depend
eavesdropper cannot be known in some situations. Because the on the CSI of legitimate users and eavesdroppers, whereas the
eavesdroppers’ CSI is unknown at the transmitters, the expres- perfect CSI is difficult to get in many situations due to the
sion of the instantaneous secrecy rate is unavailable. Therefore, limited estimation and feedback or other reasons.
the instantaneous optimization cannot be performed. Then, a The challenges stemmed from the aspect of wireless chan-
probabilistic view of security or a QoS-based optimization nels are because of the difficulties of accurate channel estima-
can be considered for secure transmission designs. Such as in tion for wiretap channels, and the considerations of channel
[256], a strategy of transmission antenna selection to enhance correlations, time varying, and node mobility. First, how to get
the secrecy performance of MIMO wiretap channels without the perfect CSI to achieve the optimal security performance
eavesdroppers’ CSI is proposed based on three important is difficult to deal with, especially when the eavesdropper is
metrics, i.e., the probability of non-zero secrecy capacity, the inactive. Furthermore, accurate channel estimation may cause
secrecy outage probability, and the ǫ-outage secrecy capacity. unacceptably high overhead in pilot frequency and power
In [47], secrecy sum rate maximization considering each user’s consumption. This is a particularly severe problem in massive
QoS constraint and unknown eavesdropper’s CSI is investi- MIMO networks as the overhead may grow rapidly with the
gated for a non-orthogonal multiple access system. In [91], a antenna number. Even worse, the process of channel estimation
QoS-based secure strategy is addressed to enhance the security may be attacked by pilot contamination attack which not only
of a cooperative relay network without eavesdropper’s CSI. It dramatically reduces the achievable secrecy capacity but is
is worth pointing out that, exploiting AN or jamming signal to also difficult to detect [17]. Second, high channel correlations
enhance secrecy has been demonstrated to be effective when have been observed in [278] even when the spatial separation
the eavesdropper’s CSI is unknown or imperfect [9]. is much larger than half-wavelength [279]. This indicates
that the spatial correlations of wireless channels may vary in
VII. D ISCUSSIONS ON F UTURE D IRECTIONS AND different environments and the half-wavelength decorrelation
C HALLENGES assumption may not always hold [279]. Therefore, the secure
It has been shown in previous sections that the physical layer transmission designs considering the channel correlations is
security has attracted increasing concerns. Some great progress also a challenging problem in future. Third, the time-varying
has been made in the fields of information-theoretical security characteristics of channels and the mobility of terminals are
and optimal secure designs at physical layer. However, it has also severe issues in physical-layer secure communications
been observed that many studies in the existing works are per- since the channel qualities may vary dramatically over time
formed with some special assumptions on CSI, eavesdropper and space. Therefore, how to simultaneously guarantee the
model, and application scenarios. These assumptions may be security, reliability, and robustness of a secure transmission
unpractical or even contrary to real conditions. Therefore, there scheme with the problems mentioned above will be challeng-
are still many significant problems needed to be investigated to ing in future work.
promote the practical applicability of physical layer security.
In the following, some possible future directions and open B. The Impacts of Adversary Model
challenges are simply discussed. Since the future work in
The impacts of attack modes and adversary models are
physical layer security is very extensive, only a few directions
also important issues for secure transmission that has not
are discussed.
yet been deeply explored. Much existing literature assumes
that the adversaries merely passively listen to the secure
A. The Influences of Wireless Channels communications. In other words, there are no collaboration
The influences of wireless channels on secrecy must be and information exchange among the adversaries. Neverthe-
further studied. In literature, it is often assumed that the less, the adversaries may actively collaborate and exchange
channels to legitimate user and eavesdropper are uncorrelated. their outputs in practice to interpret the confidential messages
The uncorrelated property is believed to be the foundation to [280]. Moreover, a slightly more sophisticated adversary may
assume that the eavesdroppers cannot estimate the channels be able to predict the channels for improving the eaves-
of legitimate transceivers. However, this assumption has its dropping qualities. Some intelligent adversaries may attempt
limitations considering some practical scenarios. For instance, to manipulate the propagation environment for strengthening
when the transceivers as well as the eavesdroppers lie in a their advantages and undermining information security [281],
insufficiently rich scattering environment as discussed in [20], [282]. When these observations discussed above are taken into
the assumption of uncorrelated channels is then impractical. account, the transmission strategy designs for physical layer
In addition, much existing literature simply assume that the security will be facing great challenges.
channels are quasi-static or even completely static. However, if The challenges in this direction can be discussed from the
the channels are somewhat dynamic, the resulting conclusions following aspects. On the one hand, the optimization and
in those works may be in conflict with the real settings. Fur- design in physical layer security will become more compli-
thermore, the relative spatial locations between the transceivers cated when hybrid attacks are imposed on wireless information
 26

transmission, such as eavesdropping attack, jamming attack, which the communication equipments may be low-cost with
denial-of-service attack, spoofing attack, message falsifica- small battery capacity, the hardware impairments may be more
tion/injection attack, etc. It will be of particular importance to severe issues for implementing physical layer security.
develop new techniques to jointly defend against hybrid wire-
less attacks [14]. On the other hand, great difficulties result D. The Joint Designs of Physical Layer Security and Classic
from the intelligent adversaries that not only can efficiently Cryptographic Security
collaborate with each other and actively manipulate propaga-
Some efforts may be needed for seeking deep insights into
tion environment for attacks, but also can autonomously learn
physical layer security and classic cryptographic security. In
the knowledge of the associated wireless network to find its
future, 5G network and beyond require ultra-strong security
weakness and then to implement adaptive attacks. Therefore, it
to support extremely secure service. Classic cryptographic
is challenging to develop well-performing secure mechanisms
security at the high cost of computational complexity, is
to defend against the intelligent adversaries.
usually deployed at the higher layers of protocol stack. As
an alternative security technology, physical layer security has
C. The Influences of Hardware Impairments the advantages of lower complexity and resource savings. Any
single security technology may not satisfy the demands of
Hardware impairments are nonnegligible factors which
high security in future. Therefore, a natural question is how to
should also be taken into account in physical layer security.
jointly exploit the advantages of the two security technologies.
So far, a great deal of works on the designs of security
Then, the cross-layer analysis and design combined with
strategies assume that the transceiver hardware is perfect.
physical layer security and classic cryptographic security come
However, hardware impairments truly exist in practice, due
naturally to mind to provide a comprehensive security solution
to nonlinear power amplifiers, in-phase and quadrature (I/Q)
from each layer of protocol stack.
imbalance, frequency and phase offsets, quantization noise,
To this end, there are many challenging problems needed
and synchronization errors [283]. For instance, I/Q imbalance
to be solved in this direction, such as the secure network
can attenuate the amplitude and rotate the phase of the desired
framework, secure coding scheme, secure network protocol,
constellation, while it can create an additional signal from
hybrid encryption algorithm, and so on. In future, the network
the mirror subcarrier which leads to a symbol error rate. In
architecture presents heterogeneous features, where the com-
the presence of nonlinearities of power amplifiers, the bit
munication nodes are deployed with dissimilar characteristics
error rate may increases remarkably compared to linear power
such as computing capacity, energy supply capacity, radio
amplifiers [284]. Although the deleterious impacts of hardware
access technologies, protocol stack architecture, etc. This
impairments on the security performance can be mitigated by
requires that the joint security strategy designs can adapt to the
calibration and compensation algorithms, residual distortions
heterogeneous architecture of networks, the variety of nodes,
at the transceivers are inevitable [283].
and the diversification of radio access technologies. This is
Many unknown challenges may be caused by hardware
significant but challenging work, since a joint security scheme
impairments in the fifth generation (5G) and beyond networks
for high level secrecy is usually followed with extremely
where novel physical layer technologies will be deployed,
high complexity which may limit its practical application.
such as the technologies of massive MIMO, mm-Wave, and
Moreover, the joint security scheme is expected to have a good
full duplex. In massive MIMO systems, additional challenges
scalability which allows the minimum amount of recomputa-
root in decreasing the hardware cost and increasing the power
tion to update protocol parameters if some components of a
efficiency on antenna array which rise to hardware impair-
network are changed. Therefore, in practice, how to design a
ments. Moreover, due to the very large size of antenna array,
simple but well-performing joint security scheme to tradeoff
standard algorithms for hardware impairment compensation,
between the performance and the complexity is an urgent need
such as digital predistortion and phase-noise estimation and
to be addressed.
compensation may be too complex in a massive MIMO system
[285]. The mm-Wave technologies utilize high frequency of
mm-Wave band, ranging from 3 ∼ 300 GHz. Due to the E. The Global Optimization with Security, Reliability, and
very small wavelength, the mm-Wave networks are different Throughput
from the conventional microwave networks in the following To achieve the optimal network performance and user
ways: large number of antennas, sensitivity to blockages, and experience in a wireless network, the security, reliability, and
variable propagation laws, which may deteriorate the harmful throughput should be considered jointly in system designs
influence of hardware impairments to secure transmissions. In [14]. However, in many existing works, these performance
full duplex systems where the information is exchanged on the metrics are taken into account individually and separately
same frequency and time slot, the residual self-interference is to reduce the difficulty in system designs. Consequently,
still remained due to the impairments of hardware interfer- the proposed security mechanisms are potentially suboptimal,
ence suppression methods, and signal processing technologies since the three factors interact with each other. For instance
are needed to be addressed to suppress the residual self- [14], the reliability and throughput of the legitimate channel
interference thoroughly. In addition to those challenges men- can be improved by increasing the transmission power which
tioned above, in some infrastructureless networks and low- however may improve the capacity of wiretap channel and
end networks (such as some specific scenarios in IoT) in increase the probability of successful eavesdropping. Likewise,
 27

although we can increase the coding rate at the transmitter VIII. C ONCLUSIONS
for improving the security level while reducing the intercept It is believed that physical layer security is a promising
probability, this leads to a decrease in transmission reliability, technology to strengthen the secrecy of confidential informa-
since higher coding rate may increase the outage probability tion delivery in many emerging wireless networks in which the
of legitimate channel. information security has not been well solved by the conven-
In order to achieve the near-perfect system performance, tional cryptographic methods. To understand the advantages
the global optimization with the joint considerations of se- of physical layer security, a comparison is first made between
curity, reliability and throughput is needed to be carried out, this security technology and the conventional cryptographic
which may be challenging and intractable. For formulating encryption. Then, the survey mainly focuses on providing
and solving such complicated multi-objective problems, some a comprehensive overview on the optimization and design
convex/nonconvex optimization techniques and game theory, of physical-layer security transmission. The typical wiretap
as well as stochastic geometry, will be widely applied in channel models are introduced to cover common scenarios
this field [26], [138]. Furthermore, the EE of a network and systems in physical layer security. The research topics
attracts increasing concerns at present and in future. When in this field are summarized from secure resource allocation,
the requirement of EE is imposed on the global optimization beamforming/precoding, and antenna/node selection and co-
discussed above, the secure transmission designs will be operation. Towards these research topics, we then discuss the
extremely complicated work which calls for innovative efforts performance metrics and fundamental optimization problems
to develop novel optimization theories and technologies. raised in the system optimization and design, which involve
the secrecy rate/capacity, secrecy outage probability/capacity,
power/energy consumption, and secure EE. The practical
F. The Commercial Application of Physical Layer Security significance and applied scenarios of the metrics are also
investigated in the survey. Each research topic of physical-
It is largely unexplored to apply the technologies of physical layer security designs involves using these performance met-
layer security into commercial wireless networks. In fact, the rics to formulate optimization problems according to specific
most research work on physical layer security still stays at application conditions. Thereafter, the state of the art of
the theory stage. The opportunities of applying physical layer optimization and design in physical layer security is reviewed
security into real commercial networks will be quite rich while from the perspectives of the aforementioned research topics.
following numerous difficulties and challenges that are from In each research topic, the great efforts are presented from
not only the technical flaws of the proposed secure strategies four categories of fundamental optimization problems, such
but also the limitations of existing network architecture and as maximization of achievable secrecy rate, minimization of
technologies, such as the hurdles from the applicability of secrecy outrage probability, minimization of power consump-
existing network framework, the expansibility of underlying tion, and maximization of secure EE. Numerous optimization
air interface, and the constraints of network resources [10]. approaches and solution schemes are investigated in the survey
Some new technical challenges will also be raised when to tackle different problems in security designs.
physical layer security are applied into the burgeoning wireless One of the major issues in the physical-layer security
networks, such as high-speed mobile networks, device to designs is the imperfect CSI problem. To achieve the optimal
device communications, cognitive radio networks, and IoT. performance of system designs, the transmitters need to know
For example, in high-speed mobile networks as representative the CSI of both the legitimate users and the eavesdroppers.
Internet of Vehicles and railway communication systems, the However, in practice, getting the perfect CSI of the eavesdrop-
rapid changes of wireless channels and terminal positions pers is very difficult or even impossible. This problem exists
require to propose fast CSI evaluation schemes and dynamic in all research topics of physical-layer security designs. In the
authentication frameworks. In device to device communica- survey, we review the existing assumptions of CSI which have
tions, due to direct communications between two mobile users been considered in physical layer security, while we discuss
without the supports of base stations or core networks, it is three ways to characterize the uncertainties of the imperfect
more difficult to establish a secure and reliable connection. eavesdropper’s CSI. It is observed that, to cope with the
Cognitive radio technique, as a promising technique to allevi- problems of the imperfect or unknown CSI of eavesdroppers,
ate spectrum scarcity, has inherent vulnerabilities in physical the robust security designs, probabilistic view of security, or
layer spectrum sensing, such as the harmful interference from QoS-based optimization is usually considered in physical layer
secondary users and the impersonation attack of disguised security to get a compromise solution. In addition, we discuss
secondary users. To detect the disguised secondary users and possible future trends and open challenges from the aspects
to mitigate secondary interference, the terminals in cognitive involving the problems of imperfect CSI, eavesdropper mod-
radio networks should have the ability of autonomous learning. els, and hardware impairments, as well as cross-layer security
Machine learning is a powerful tool that can bring inspirations designs, global performance optimizations, and commercial
to cope with the potential challenges. IoT has a lot of particular application of physical layer security.
characteristics, such as a massive number of devices, low-cost
hardware, limited battery capacity, weak computation ability, R EFERENCES
and distinct service scenarios, all of which bring unprece- [1] J. Cao, M. Ma, H. Li, Y. Zhang, and Z. Luo, “A survey on security
dented challenges in implementing physical layer security. aspects for LTE and LTE-A networks,” IEEE Commun. Surveys Tuts.,
 28

vol. 16, no. 1, pp. 283–302, First Quarter 2014. precoder design.” IEEE Trans. Inf. Forensics Security, vol. 6, no. 3,
[2] W. Stallings, Cryptography and network security: principles and prac- pp. 640–649, Sep. 2011.
tice. New York, USA: Pearson Education, Inc., 2011. [27] Z. Kong, S. Yang, F. Wu, and S. Peng, “Iterative distributed minimum
[3] Y. Liang, H. V. Poor, and S. Shamai, “Information theoretic security,” total-MSE approach for secure communications in MIMO interference
Found. Trends in Commun. Inf. Theory, vol. 5, no. 4-5, pp. 355–580, channels,” IEEE Trans. Inf. Forensics Security, vol. 11, no. 3, pp. 594–
Apr. 2009. 608, Mar. 2016.
[4] C. E. Shanon, “Communication theory of secrecy systems,” Bell Labs [28] J. Huang and A. L. Swindlehurst, “Cooperative jamming for secure
Technical Journal, vol. 28, no. 4, pp. 656–715, Apr. 1949. communications in MIMO relay networks,” IEEE Trans. Signal Pro-
[5] A. D. Wyner, “The wire-tap channel,” Bell Sys. Tech. J., vol. 54, no. 8, cess., vol. 59, no. 10, pp. 4871–4884, Oct. 2011.
pp. 1355–1387, Oct. 1975. [29] J. Mo, M. Tao, Y. Liu, and R. Wang, “Secure beamforming for MIMO
[6] I. Csiszár and J. Körner, “Broadcast channels with confidential mes- two-way communications with an untrusted relay,” IEEE Trans. Signal
sages,” IEEE Trans. Inf. Theory, vol. 24, no. 3, pp. 339–348, May Process., vol. 62, no. 9, pp. 2185–2199, May 2014.
1978. [30] Q. Li, Y. Yang, W. K. Ma, M. Lin, J. Ge, and J. Lin, “Robust
[7] S. K. Leung-Yan-Cheong and M. E. Hellman, “The Gaussian wire-tap cooperative beamforming and artificial noise design for physical-layer
channel,” IEEE Trans. Inf. Theory, vol. IT-24, no. 4, pp. 451–456, Jul. secrecy in AF multi-antenna multi-relay networks,” IEEE Trans. Signal
1978. Process., vol. 63, no. 1, pp. 206–220, Jan. 2015.
[8] R. Bassily, E. Ekrem, X. He, E. Tekin, M. R. B. J. Xie, S. Ulukus, [31] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge,
and A. Yener, “Cooperative security at the physical layer: A summary U.K.: Cambridge Univ. Press, 2004.
of recent advances,” IEEE Signal Process. Mag., vol. 30, no. 5, pp. [32] G. Song and Y. Li, “Utility-based resource allocation and scheduling
16–28, Sep. 2013. in OFDM-Based wireless broadband networks,” IEEE Commun. Mag.,
[9] Y. W. P. Hong and C. C. J. Kuo, “Enhancing physical-layer secrecy vol. 43, no. 12, pp. 127–134, Dec. 2005.
in multiantenna wireless systems: An overview of signal processing [33] B. Bai, W. Chen, and Z. Cao, “Outage optimal subcarrier allocation
approaches,” IEEE Signal Process. Mag., vol. 30, no. 5, pp. 29–40, for downlink secure OFDMA systems,” in Proc. IEEE Global Com-
Sep. 2013. munications Conference (GLOBECOM) Workshops, Austin, TX, Dec.
[10] A. Mukherjee, S. A. A. Fakoorian, J. Huang, and A. L. Swindlehurst, 2014, pp. 1320–1325.
“Principles of physical layer security in multiuser wireless networks: A [34] A. Jindal and R. Bose, “Resource allocation for secure multicarrier
survey,” IEEE Commun. Surveys Tuts., vol. 16, no. 3, pp. 1550–1573, AF relay system under total power constraint,” IEEE Commun. Lett.,
Third Quarter 2014. vol. 19, no. 2, pp. 231–234, Feb. 2015.
[11] A. Yener and S. Ulukus, “Wireless physical-layer security: Lessons [35] X. Wang, M. Tao, J. Mo, and Y. Xu, “Power and subcarrier allocation
learned from information theory,” Proceedings of the IEEE, vol. 103, for physical-layer security in OFDMA-based broadband wireless net-
no. 10, pp. 1814–1825, Oct. 2015. works,” IEEE Trans. Inf. Forensics Security, vol. 6, no. 3, pp. 693–702,
[12] H.-M. Wang and X.-G. Xia, “Enhancing wireless secrecy via cooper- Sep. 2011.
ation: Signal design and optimization,” IEEE Commun. Mag., vol. 53, [36] D. W. K. Ng, E. S. Lo, and R. Schober, “Secure resource allocation
no. 12, pp. 47–53, Dec. 2015. and scheduling for OFDMA decode-and-forward relay networks,” IEEE
Trans. Wireless Commun., vol. 10, no. 10, pp. 3528–3540, Oct. 2011.
[13] Y. Liu, H. H. Chen, and L. Wang, “Physical layer security for next
[37] I. Krikidis, J. S. Thompson, P. M. Grant, and S. Mclaughlin, “Power
generation wireless networks: Theories, technologies, and challenges,”
allocation for cooperative-based jamming in wireless networks with se-
IEEE Commun. Surveys Tuts., vol. 19, no. 1, pp. 347–376, First Quarter
crecy constraints,” in Proc. IEEE Global Communications Conference
2017.
(GLOBECOM) Workshops, Miami, FL, USA, Dec. 2010, pp. 1177–
[14] Y. Zou, J. Zhu, X. Wang, and L. Hanzo, “A survey on wireless security:
1181.
Technical challenges, recent advances and future trends,” Proceedings
[38] C. Jeong and I. M. Kim, “Optimal power allocation for secure mul-
of the IEEE, vol. 104, no. 9, pp. 1727–1765, Sep. 2016.
ticarrier relay systems,” IEEE Trans. Signal Process., vol. 59, no. 11,
[15] A. Mukherjee, “Physical-layer security in the internet of things: Sens- pp. 5428–5442, Nov. 2011.
ing and communication confidentiality under resource constraints,” [39] J. Huang and A. L. Swindlehurst, “Robust secure transmission in
Proceedings of the IEEE, vol. 103, no. 10, pp. 1747–1761, Oct 2015. MISO channels based on worst-case optimization,” IEEE Trans. Signal
[16] L. Zhou, D. Wu, B. Zheng, and M. Guizani, “Joint physical-application Process., vol. 60, no. 4, pp. 1696–1707, Apr. 2012.
layer security for wireless multimedia delivery,” IEEE Commun. Mag., [40] K. H. Park, T. Wang, and M. S. Alouini, “On the jamming power
vol. 52, no. 3, pp. 66–72, Mar. 2014. allocation for secure amplify-and-forward relaying via cooperative
[17] D. Kapetanovic, G. Zheng, and F. Rusek, “Physical layer security for jamming,” IEEE J. Sel. Areas Commun., vol. 31, no. 9, pp. 1741–1750,
massive MIMO: An overview on passive eavesdropping and active Sep. 2013.
attacks,” IEEE Commun. Mag., vol. 53, no. 6, pp. 21–27, Jun. 2015. [41] L. Wang, M. Elkashlan, J. Huang, N. H. Tran, and T. Q. Duong,
[18] L. J. Rodriguez, N. H. Tran, T. Q. Duong, and T. Le-Ngoc, “Physical “Secure transmission with optimal power allocation in untrusted relay
layer security in wireless cooperative relay networks: State of the art networks,” IEEE Wireless Commun. Lett., vol. 3, no. 3, pp. 289–292,
and beyond,” IEEE Commun. Mag., vol. 53, no. 12, pp. 32–39, Dec. Mar. 2014.
2015. [42] S. H. Tsai and H. V. Poor, “Power allocation for artificial-noise secure
[19] N. Yang, L. Wang, G. Geraci, and M. Elkashlan, “Safeguarding 5g MIMO precoding systems,” IEEE Trans. Signal Process., vol. 62,
wireless communication networks using physical layer security,” IEEE no. 13, pp. 3479–3493, July 2014.
Commun. Mag., vol. 53, no. 4, pp. 20–27, Apr. 2015. [43] J. H. Lee, “Full-duplex relay for enhancing physical layer security in
[20] W. Trappe, “The challenges facing physical layer security,” IEEE multi-hop relaying systems,” IEEE Commun. Lett., vol. 19, no. 4, pp.
Commun. Mag., vol. 53, no. 6, pp. 16–20, Jun. 2015. 525–528, Apr. 2015.
[21] Y. S. Shiu, S. Y. Chang, H. C. Wu, C. H. Huang, and H. H. [44] ——, “Optimal power allocation for physical layer security in multi-
Chen, “Physical layer security in wireless networks: A tutorial,” IEEE hop DF relay networks,” IEEE Trans. Wireless Commun., vol. 15, no. 1,
Wireless Commun., vol. 18, no. 2, pp. 66–74, Apr. 2011. pp. 28–38, Jan. 2016.
[22] Y. Zou, J. Zhu, X. Wang, and V. Leung, “Improving physical-layer [45] T. X. Zheng and H.-M. Wang, “Optimal power allocation for artificial
security in wireless communications using diversity techniques,” IEEE noise under imperfect CSI against spatially random eavesdroppers,”
Netw., vol. 29, no. 1, pp. 42–48, Jan.-Feb. 2015. IEEE Trans. Veh. Technol., vol. 65, no. 10, pp. 8812–8817, Oct. 2016.
[23] E. Ekrem and S. Ulukus, “Secure broadcasting using multiple anten- [46] A. Benfarah, S. Tomasin, and N. Laurenti, “Power allocation in
nas,” Journal of Communications and Networks, vol. 12, no. 5, pp. multiuser parallel Gaussian broadcast channels with common and
411–432, Oct. 2010. confidential messages,” IEEE Trans. Commun., vol. 64, no. 6, pp.
[24] H. Lee, C. Song, J. Moon, and I. Lee, “Precoder designs for 2326–2339, Jun. 2016.
MIMO Gaussian multiple access wiretap channels,” IEEE Trans. Veh. [47] Y. Zhang, H.-M. Wang, Q. Yang, and Z. Ding, “Secrecy sum rate max-
Technol., vol. 66, no. 9, pp. 8563–8568, Sep. 2017. imization in non-orthogonal multiple access,” IEEE Wireless Commun.
[25] E. Tekin and A. Yener, “The Gaussian multiple access wire-tap Lett., vol. 20, no. 5, pp. 930–933, May 2016.
channel,” IEEE Trans. Inf. Theory, vol. 54, no. 12, pp. 5747–5755, [48] J. Chen, X. Chen, W. H. Gerstacker, and D. W. K. Ng, “Resource
Dec. 2008. allocation for a massive MIMO relay aided secure communication,”
[26] S. A. A. Fakoorian and A. L. Swindlehurst, “MIMO interference IEEE Trans. Inf. Forensics Security, vol. 11, no. 8, pp. 1700–1711,
channel with confidential messages: Achievable secrecy rates and Aug. 2016.
 29

[49] H. Zhang, H. Xing, J. Cheng, A. Nallanathan, and V. Leung, “Se- [71] H. Zhang, Y. Huang, S. Li, and L. Yang, “Energy-efficient precoder
cure resource allocation for OFDMA two-way relay wireless sensor design for MIMO wiretap channels,” IEEE Commun. Lett., vol. 18,
networks without and with cooperative jamming,” IEEE Trans. Ind. no. 9, pp. 1559–1562, Sep. 2014.
Informat., vol. 12, no. 5, pp. 1714–1725, Oct. 2016. [72] R. Liu and H. V. Poor, “Secrecy capacity region of a multiple-antenna
[50] S. Karachontzitis, S. Timotheou, I. Krikidis, and K. Berberidis, Gaussian broadcast channel with confidential messages,” IEEE Trans.
“Security-aware max-min resource allocation in multiuser OFDMA Inf. Theory, vol. 55, no. 3, pp. 1235–1249, Mar. 2009.
downlink,” IEEE Trans. Inf. Forensics Security, vol. 10, no. 3, pp. [73] R. Liu, T. Liu, H. V. Poor, and S. Shamai, “Multiple-input multiple-
529–542, Mar. 2015. output Gaussian broadcast channels with confidential messages,” IEEE
[51] D. W. K. Ng, E. S. Lo, and R. Schober, “Energy-efficient resource Trans. Inf. Theory, vol. 56, no. 9, pp. 4215–4227, Sep. 2010.
allocation for secure OFDMA systems,” IEEE Trans. Veh. Technol., [74] S. A. A. Fakoorian and A. L. Swindlehurst, “Dirty paper coding
vol. 61, no. 6, pp. 2572–2585, Jul. 2012. versus linear GSVD-based precoding in MIMO broadcast channel
[52] C. Jeong, I. M. Kim, and D. I. Kim, “Joint secure beamforming design with confidential messages,” in Proc. IEEE Global Communications
at the source and the relay for an amplify-and-forward MIMO untrusted Conference (GLOBECOM), Houston, TX, Dec. 2011.
relay system,” IEEE Trans. Signal Process., vol. 60, no. 1, pp. 310–325, [75] Khisti, A., Wornell, and W. G., “Secure transmission with multiple
Jan. 2012. antennas I: The MISOME wiretap channel,” IEEE Trans. Inf. Theory,
[53] Z. Liu, C. Chen, L. Bai, and H. Xiang, “Transmit power minimization vol. 56, no. 7, pp. 3088–3104, Jul. 2010.
beamforming via amplify-and-forward relays in wireless networks with [76] ——, “Secure transmission with multiple antennas II: The MIMOME
multiple eavesdroppers,” in Proc. IEEE International Conference on wiretap channel,” IEEE Trans. Inf. Theory, vol. 56, no. 11, pp. 5515–
Communications (ICC), Sydney, NSW, Australia, Jun. 2014, pp. 4698– 5532, Nov. 2010.
4703. [77] S. A. A. Fakoorian and A. L. Swindlehurst, “Optimal power allocation
for GSVD-based beamforming in the MIMO Gaussian wiretap chan-
[54] Q. Shi, W. Xu, J. Wu, E. Song, and Y. Wang, “Secure beamforming
nel,” in Proc. IEEE International Symposium on Information Theory
for MIMO broadcasting with wireless information and power transfer,”
(ISIT), Cambridge, MA, USA, Jul. 2012, pp. 2321–2325.
IEEE Trans. Wireless Commun., vol. 14, no. 5, pp. 2841–2853, May
[78] G. Geraci, M. Egan, J. Yuan, and A. Razi, “Secrecy sum-rates for multi-
2015.
user MIMO regularized channel inversion precoding,” IEEE Trans.
[55] T. Lv, H. Gao, and S. Yang, “Secrecy transmit beamforming for Commun., vol. 60, no. 11, pp. 3472 – 3482, Nov. 2012.
heterogeneous networks,” IEEE J. Sel. Areas Commun., vol. 33, pp. [79] G. Geraci, H. S. Dhillon, J. G. Andrews, J. Yuan, and I. B. Collings,
1154–1170, Jun. 2015. “Physical layer security in downlink multi-antenna cellular networks,”
[56] P. Zhao, M. Zhang, H. Yu, H. Luo, and W. Chen, “Robust beamforming IEEE Trans. Commun., vol. 62, no. 6, pp. 2006 – 2021, Jun. 2014.
design for sum secrecy rate optimization in MU-MISO networks,” [80] G. Geraci, R. Couillet, J. Yuan, M. Debbah, and I. B. Collings, “Large
IEEE Trans. Inf. Forensics Security, vol. 10, no. 9, pp. 1812–1823, system analysis of linear precoding in MISO broadcast channels with
Sep. 2015. confidential messages,” IEEE J. Sel. Areas Commun., vol. 31, no. 9,
[57] G. Zheng, P. Arapoglou, and B. Ottersten, “Physical layer security in pp. 1660–1671, Sep. 2013.
multibeam satellite systems,” IEEE Trans. Wireless Commun., vol. 11, [81] N. Yang, G. Geraci, J. Yuan, and R. Malaney, “Confidential broad-
no. 2, pp. 852–863, Feb. 2012. casting via linear precoding in non-homogeneous MIMO multiuser
[58] X. Wang, K. Wang, and X. D. Zhang, “Secure relay beamforming networks,” IEEE Trans. Commun., vol. 62, no. 7, pp. 2515–2530, Jul.
with imperfect channel side information,” IEEE Trans. Veh. Technol., 2014.
vol. 62, no. 5, pp. 2140–2155, Jun. 2013. [82] B. He, N. Yang, X. Zhou, and J. Yuan, “Base station cooperation for
[59] L. Dong, Z. Han, A. P. Petropulu, and H. V. Poor, “Improving wireless confidential broadcasting in multi-cell networks,” IEEE Trans. Wireless
physical layer security via cooperating relays,” IEEE Trans. Signal Commun., vol. 14, no. 10, pp. 5287–5299, Oct. 2015.
Process., vol. 58, no. 3, pp. 1875–1888, Mar. 2010. [83] Q. Li, M. Hong, H. T. Wai, Y. F. Liu, W. K. Ma, and Z. Q. Luo,
[60] H.-M. Wang, Q. Yin, and X. G. Xia, “Distributed beamforming for “Transmit solutions for MIMO wiretap channels using alternating
physical-layer security of two-way relay networks,” IEEE Trans. Signal optimization,” IEEE J. Sel. Areas Commun., vol. 31, no. 9, pp. 1714–
Process., vol. 60, no. 7, pp. 3532–3545, Jul. 2012. 1727, Sep. 2013.
[61] Y. Yang, Q. Li, W. K. Ma, J. Ge, and P. C. Ching, “Cooperative secure [84] W. C. Liao, T. H. Chang, W. K. Ma, and C. Y. Chi, “QoS-based transmit
beamforming for AF relay networks with multiple eavesdroppers,” beamforming in the presence of eavesdroppers: An optimized artificial-
IEEE Signal Process. Lett., vol. 20, no. 1, pp. 35–38, Jan. 2013. noise-aided approach,” IEEE Trans. Signal Process., vol. 59, no. 3, pp.
[62] X. Zhang, X. Zhou, and M. R. Mckay, “On the design of artificial- 1202–1216, Mar. 2011.
noise-aided secure multi-antenna transmission in slow fading channels,” [85] P. H. Lin, S. H. Lai, S. C. Lin, and H. J. Su, “On secrecy rate of
IEEE Trans. Veh. Technol., vol. 62, no. 5, pp. 2170 – 2181, Jun. 2013. the generalized artificial-noise assisted secure beamforming for wiretap
[63] J. Zhu, R. Schober, and V. K. Bhargava, “Secure transmission in multi- channels,” IEEE J. Sel. Areas Commun., vol. 31, no. 9, pp. 1728–1740,
cell massive MIMO systems,” IEEE Trans. Wireless Commun., vol. 13, Sep. 2012.
no. 9, pp. 1286–1291, Sep. 2014. [86] N. Romero-Zurita, M. Ghogho, and D. Mclernon, “Outage probability
[64] D. Wang, B. Bai, W. Chen, and Z. Han, “Energy efficient secure based power distribution between data and artificial noise for physical
communication over decode-and-forward relay channels,” IEEE Trans. layer security,” IEEE Signal Process. Lett., vol. 19, no. 2, pp. 71–74,
Commun., vol. 63, no. 3, pp. 892–905, Mar. 2015. Feb. 2012.
[87] H.-M. Wang, C. Wang, and W. Ng, “Artificial noise assisted secure
[65] C. R. Murthy, A. K. Jagannatham, and B. D. Rao, “Training-based
transmission under training and feedback,” IEEE Trans. Signal Pro-
and semiblind channel estimation for MIMO systems with maximum
cess., vol. 63, no. 23, pp. 6285–6298, Dec. 2015.
ratio transmission,” IEEE Trans. Signal Process., vol. 54, no. 7, pp.
[88] B. Wang, P. Mu, and Z. Li, “Secrecy rate maximization with artificial-
2546–2558, Jul. 2006.
noise-aided beamforming for MISO wiretap channels under secrecy
[66] S. A. A. Fakoorian and A. L. Swindlehurst, “On the optimality of linear outage constraint,” IEEE Commun. Lett., vol. 19, no. 1, pp. 18–21,
precoding for secrecy in the MIMO broadcast channel,” IEEE J. Sel. Jan. 2015.
Areas Commun., vol. 31, no. 9, pp. 1701–1713, Sep. 2013. [89] H.-M. Wang, T. Zheng, and X.-G. Xia, “Secure MISO wiretap channels
[67] H. Long, W. Xiang, Y. Zhang, and Y. Liu, “Secrecy capacity enhance- with multiantenna passive eavesdropper: Artificial noise vs. artificial
ment with distributed precoding in multirelay wiretap systems,” IEEE fast fading,” IEEE Trans. Wireless Commun., vol. 14, no. 1, pp. 94–
Trans. Inf. Forensics Secur., vol. 8, no. 1, pp. 229–238, Jan. 2013. 106, Jan. 2015.
[68] M. F. Hanif, L. N. Tran, M. Juntti, and S. Glisic, “On linear precoding [90] N. Yang, M. Elkashlan, T. Q. Duong, and J. Yuan, “Optimal transmis-
strategies for secrecy rate maximization in multiuser multiantenna sion with artificial noise in MISOME wiretap channels,” IEEE Trans.
wireless networks,” IEEE Trans. Signal Process., vol. 62, no. 14, pp. Veh. Technol., vol. 65, no. 4, pp. 2170–2181, Apr. 2015.
3536–3551, Jul. 2014. [91] H.-M. Wang, M. Luo, X.-G. Xia, and Q. Yin, “Joint cooperative
[69] C. H. Lin, S. H. Tsai, and Y. P. Lin, “Secure transmission using MIMO beamforming and jamming to secure AF relay systems with individual
precoding,” IEEE Trans. Inf. Forensics Security, vol. 9, no. 5, pp. 801– power constraint and no eavesdropper’s CSI,” IEEE Signal Process.
813, May 2014. Letters, vol. 20, no. 1, pp. 39–42, Jan. 2013.
[70] F. Wu, R. Zhang, L. L. Yang, and W. Wang, “Transmitter precoding- [92] S. Goel and R. Negi, “Guaranteeing secrecy using artificial noise,”
aided spatial modulation for secrecy communications,” IEEE Trans. IEEE Trans. Wireless Commun., vol. 7, no. 6, pp. 2180–2189, Jun.
Veh. Technol., vol. 65, no. 1, pp. 467–471, Jan. 2016. 2008.
 30

[93] M. F. Hanif, “Efficient algorithm for selecting secrecy rate maximizing [116] E. Tekin and A. Yener, “The general Gaussian multiple-access and
antennas,” IEEE Commun. Lett., vol. 17, no. 9, pp. 1818–1821, Sep. two-way wiretap channels: Achievable rates and cooperative jamming,”
2013. IEEE Trans. Inf. Theory, vol. 54, no. 6, pp. 2735–2751, Jun. 2008.
[94] N. Yang, P. L. Yeoh, M. Elkashlan, R. Schober, and I. B. Collings, [117] J. Yang, I. M. Kim, and I. K. Dong, “Optimal cooperative jamming for
“Transmit antenna selection for security enhancement in MIMO wire- multiuser broadcast channel with multiple eavesdroppers,” IEEE Trans.
tap channels,” IEEE Trans. Commun., vol. 61, no. 1, pp. 144–154, Jan. Wireless Commun., vol. 12, no. 6, pp. 2840–2852, Jun. 2013.
2013. [118] J. H. Lee and C. Wan, “Multiuser diversity for secrecy communications
[95] S. Yan, N. Yang, R. Malaney, and J. Yuan, “Transmit antenna selection using opportunistic jammer selection: Secure DoF and jammer scaling
with Alamouti coding and power allocation in MIMO wiretap chan- law,” IEEE Trans. Signal Process., vol. 62, no. 4, pp. 828–839, Feb.
nels,” IEEE Trans. Wireless Commun., vol. 13, no. 3, pp. 1656–1667, 2014.
Mar. 2014. [119] C. Wang, H.-M. Wang, X.-G. Xia, and C. Liu, “Uncoordinated jam-
[96] H. Alves, R. D. Souza, M. Debbah, and M. Bennis, “Performance mer selection for securing SIMOME wiretap channels: A stochastic
of transmit antenna selection physical layer security schemes,” IEEE geometry approach,” IEEE Trans. Wireless Commun., vol. 14, no. 5,
Signal Process. Lett., vol. 19, no. 6, pp. 372–375, Jun. 2012. pp. 2596–2612, May 2015.
[97] N. Yang, P. L. Yeoh, M. Elkashlan, R. Schober, and J. Yuan, “MIMO [120] C. Wang and H.-M. Wang, “Opportunistic jamming for enhancing
wiretap channels: Secure transmission using transmit antenna selection security: Stochastic geometry modeling and analysis,” IEEE Trans. Veh.
and receive generalized selection combining,” IEEE Commun. Lett., Technol., vol. 65, no. 12, pp. 10 213–10 217, Dec. 2016.
vol. 17, no. 9, pp. 1754–1757, Sep. 2013. [121] X. He and A. Yener, “Cooperation with an untrusted relay: A secrecy
[98] N. Yang, H. A. Suraweera, I. B. Collings, and C. Yuen, “Physical perspective,” IEEE Trans. Inf. Theory, vol. 56, no. 8, pp. 3807–3827,
layer security of TAS/MRC with antenna correlation,” IEEE Trans. Aug. 2010.
Inf. Forensics Security, vol. 8, no. 1, pp. 254–259, Jan. 2013. [122] ——, “The role of an untrusted relay in secret communication,” in
[99] L. Wang, M. Elkashlan, J. Huang, R. Schober, and R. K. Mallik, Proc. IEEE International Symposium on Information Theory (ISIT),
“Secure transmission with antenna selection in MIMO Nakagami-m Toronto, Canada, Jul. 2008, pp. 2212–2216.
fading channels,” IEEE Trans. Wireless Commun., vol. 13, no. 11, pp. [123] R. Zhang, L. Song, Z. Han, and B. Jiao, “Physical layer security for
6054–6067, Nov. 2014. two-way untrusted relaying with friendly jammers,” IEEE Trans. Veh.
[100] Y. Hu and X. Tao, “Secrecy outage on transmit antenna selection with Technol., vol. 61, no. 8, pp. 3693–3704, Oct. 2012.
weighting errors at maximal-ratio combiners,” IEEE Commun. Lett., [124] J. Huang and A. L. Swindlehurst, “Joint transmit design and node
vol. 19, no. 4, pp. 597–600, Apr. 2015. selection for one-way and two-way untrusted relay channels,” in Proc.
[101] S. N. Diggavi, N. Al-Dhahir, A. Stamoulis, and A. R. Calderbank, 2013 Asilomar Conference on Signals, Systems and Computers, Pacific
“Great expectations: the value of spatial diversity in wireless networks,” Grove, Nov. 2013, pp. 1555–1559.
Proceedings of the IEEE, vol. 92, no. 2, pp. 219–270, Feb. 2004. [125] A. Mukherjee, “Imbalanced beamforming by a multi-antenna source for
[102] S. I. Kim, I. M. Kim, and J. Heo, “Secure transmission for multiuser secure utilization of an untrusted relay,” IEEE Commun. Lett., vol. 17,
relay networks,” IEEE Trans. Wireless Commun., vol. 14, no. 7, pp. no. 7, pp. 1309–1312, Jul. 2013.
3724–3737, Jul. 2015.
[126] H. Khodakarami and F. Lahouti, “Link adaptation with untrusted relay
[103] Y. Zou, X. Wang, and W. Shen, “Physical-layer security with mul-
assignment: Design and performance analysis,” IEEE Trans. Commun.,
tiuser scheduling in cognitive radio networks,” IEEE Trans. Commun.,
vol. 61, no. 12, pp. 4874–4883, Dec. 2013.
vol. 61, no. 12, pp. 5103–5113, Dec. 2013.
[127] J. Huang, A. Mukherjee, and A. L. Swindlehurst, “Secure communica-
[104] Y. Zou, X. Li, and Y. C. Liang, “Secrecy outage and diversity analysis
tion via an untrusted non-regenerative relay in fading channels,” IEEE
of cognitive radio systems,” IEEE J. Sel. Areas Commun., vol. 32,
Trans. Signal Process., vol. 61, no. 10, pp. 2536–2550, May 2013.
no. 11, pp. 2222–2236, Nov. 2014.
[128] J. Richter, C. Scheunert, S. Engelmann, and E. A. Jorswieck, “Weak
[105] D. Christopoulos, S. Chatzinotas, and B. Ottersten, “Multicast multi-
group precoding and user scheduling for frame-based satellite commu- secrecy in the multi-way untrusted relay channel with compute-and-
forward,” IEEE Trans. Inf. Forensics Security, vol. 10, no. 6, pp. 1262–
nications,” IEEE Trans. Wireless Commun., vol. 14, no. 9, pp. 4695–
1273, Jun. 2015.
4707, Sep. 2015.
[106] T. M. Hoang, T. Q. Duong, H. A. Suraweera, C. Tellambura, and H. V. [129] H. Deng, H. M. Wang, W. Guo, and W. Wang, “Secrecy transmission
Poor, “Cooperative beamforming and user selection for improving the with a helper: To relay or to jam,” IEEE Trans. Inf. Forensics Security,
security of relay-aided systems,” IEEE Trans. Commun., vol. 63, no. 12, vol. 10, no. 2, pp. 293–307, Feb. 2015.
pp. 5039–5051, Dec. 2015. [130] X. He and A. Yener, “Two-hop secure communication using an
[107] A. Mukherjee and A. L. Swindlehurst, “User selection in multiuser untrusted relay: A case for cooperative jamming,” in Proc. IEEE Global
MIMO systems with secrecy considerations,” in Proc. 2009 Asilomar Communications Conference (GLOBECOM), New Orleans, LA, Nov.
Conference on Signals, Systems and Computers, Pacific Grove, USA, 2008.
Nov. 2009, pp. 1479–1482. [131] L. Sun, T. Zhang, Y. Li, and H. Niu, “Performance study of two-hop
[108] M. Pei, A. L. Swindlehurst, D. Ma, and J. Wei, “On ergodic secrecy rate amplify-and-forward systems with untrustworthy relay nodes,” IEEE
for MISO wiretap broadcast channels with opportunistic scheduling,” Trans. Veh. Technol., vol. 61, no. 8, pp. 3801–3807, Oct. 2012.
IEEE Commun. Lett., vol. 18, no. 1, pp. 50–53, Jan. 2014. [132] I. Krikidis, J. S. Thompson, and S. Mclaughlin, “Relay selection for
[109] Y. Zou, X. Wang, and W. Shen, “Optimal relay selection for physical- secure cooperative networks with jamming,” IEEE Trans. Wireless
layer security in cooperative wireless networks,” IEEE J. Sel. Areas Commun., vol. 8, no. 10, pp. 5003–5011, Feb. 2009.
Commun., vol. 31, no. 10, pp. 2099–2111, Oct. 2013. [133] J. Chen, R. Zhang, L. Song, Z. Han, and B. Jiao, “Joint relay and
[110] V. N. Q. Bao, N. Linh-Trung, and M. Debbah, “Relay selection jammer selection for secure two-way relay networks,” IEEE Trans.
schemes for dual-hop networks under security constraints with multiple Inf. Forensics Security, vol. 7, no. 1, pp. 310–320, Feb. 2011.
eavesdroppers,” IEEE Trans. Wireless Commun., vol. 12, no. 12, pp. [134] N. E. Wu and H. J. Li, “Effect of feedback delay on secure cooper-
6076–6085, Dec. 2013. ative networks with joint relay and jammer selection,” IEEE Wireless
[111] F. S. Al-Qahtani, C. Zhong, and H. M. Alnuweiri, “Opportunistic Commun. Lett., vol. 2, no. 4, pp. 415–418, Aug. 2013.
relay selection for secrecy enhancement in cooperative networks,” IEEE [135] Y. Liu, L. Wang, T. T. Duy, M. Elkashlan, and T. Q. Duong, “Relay
Trans. Commun., vol. 63, no. 5, pp. 1756–1770, May 2015. selection for security enhancement in cognitive relay networks,” IEEE
[112] W. Wang, K. Teh, and K. Li, “Generalized relay selection for improved Wireless Commun. Lett., vol. 4, no. 1, pp. 46–49, Feb. 2015.
security in cooperative DF relay networks,” IEEE Wireless Commun. [136] H. Hui, A. L. Swindlehurst, G. Li, and J. Liang, “Secure relay and
Lett., vol. 5, no. 1, pp. 28–31, Feb. 2015. jammer selection for physical layer security,” IEEE Signal Process.
[113] L. Fan, X. Lei, T. Q. Duong, and M. Elkashlan, “Secure multiuser com- Lett., vol. 22, no. 8, pp. 1147–1151, Aug. 2015.
munications in multiple amplify-and-forward relay networks,” IEEE [137] L. Wang, Y. Cai, Y. Zou, W. Yang, and L. Hanzo, “Joint relay and
Trans. Commun., vol. 62, no. 9, pp. 3299–3310, Sep. 2014. jammer selection improves the physical layer security in the face of
[114] N. Nomikos, A. Nieto, and P. Makris et al., “Relay selection for secure CSI feedback delays,” IEEE Trans. Veh. Technol., vol. 65, no. 8, pp.
5G green communications,” Telecommunication Systems, vol. 59, no. 1, 6259–6274, Aug. 2016.
pp. 169–187, Jan. 2015. [138] N. Zhang, N. Cheng, N. Lu, X. Zhang, J. W. Mark, and X. Shen,
[115] D. Wang, B. Bai, W. Chen, and Z. Han, “Achieving high energy “Partner selection and incentive mechanism for physical layer security,”
efficiency and physical-layer security in AF relaying,” IEEE Trans. IEEE Trans. Wireless Commun., vol. 14, no. 8, pp. 4265–4276, Aug.
Wireless Commun., vol. 15, no. 1, pp. 740–752, Jan. 2016. 2015.
 31

[139] W. Y. Luo, L. Jin, K. Z. Huang, and Z. Zhong, “User selection [162] O. Gungor, J. Tan, C. E. Koksal, H. El-Gamal, and N. B. Shroff,
and resource allocation for secure multiuser MISO-OFDMA systems,” “Secrecy outage capacity of fading channels,” IEEE Trans. Inf. Theory,
Electronics Letters, vol. 47, no. 15, pp. 884–886, Jul. 2011. vol. 59, no. 9, pp. 5379–5397, Sep. 2013.
[140] X. Zhu, B. Yang, C. Chen, and L. Xue, “Cross-layer scheduling [163] J. Chen, X. Chen, T. Liu, and L. Lei, “Energy-efficient power allocation
for OFDMA-based cognitive radio systems with delay and security for secure communications in large-scale MIMO relaying systems,”
constraints,” IEEE Trans. Veh. Technol., vol. 64, no. 12, pp. 5919– in Proc. IEEE/CIC International Conference on Communications in
5934, Dec. 2015. China (ICCC), Shanghai, China, Oct. 2014, pp. 385–390.
[141] M. Hanif, M. Juntti, and L. N. Tran, “Antenna selection with erroneous [164] S. Cui, A. J. Goldsmith, and A. Bahai, “Energy-efficiency of MIMO
covariance matrices under secrecy constraints,” IEEE Trans. Veh. and cooperative MIMO techniques in sensor networks,” IEEE J. Sel.
Technol., vol. 65, no. 1, pp. 414–420, Jan. 2016. Areas Commun., vol. 22, no. 6, pp. 1089–1098, Aug. 2004.
[142] J. Kim, A. Ikhlef, and R. Schober, “Combined relay selection and [165] C. Jiang and L. J. Cimini, “Antenna selection for energy-efficient
cooperative beamforming for physical layer security,” Journal of Com- MIMO transmission,” IEEE Commun. Lett., vol. 1, no. 6, pp. 577–
munications and Networks, vol. 14, no. 4, pp. 364–373, Aug. 2012. 580, Dec. 2012.
[143] H.-M. Wang, M. Luo, Q. Yin, and X.-G. Xia, “Hybrid cooperative [166] A. G. Marques, G. B. Giannakis, F. F. Digham, and F. J. Ramos,
beamforming and jamming for physical-layer security of two-way relay “Power-efficient wireless OFDMA using limited-rate feedback,” IEEE
networks,” IEEE Trans. Inf. Forensics Security, vol. 8, no. 12, pp. 2007– Trans. Wireless Commun., vol. 7, no. 2, pp. 685–696, Feb. 2008.
2020, Dec. 2013. [167] J. Joung and S. Sun, “Power efficient resource allocation for downlink
[144] C. Wang, H.-M. Wang, and X.-G. Xia, “Hybrid opportunistic relaying OFDMA relay cellular networks,” IEEE Trans. Signal Process., vol. 60,
and jamming with power allocation for secure cooperative networks,” no. 5, pp. 2447–2459, May 2012.
IEEE Trans. Wireless Commun., vol. 14, no. 2, pp. 589–605, Feb. 2015. [168] C. Comaniciu, H. V. Poor, and R. Zhang, “An information theoretic
[145] I. Krikidis and B. Ottersten, “Secrecy sum-rate for orthogonal random framework for energy efficient secrecy,” in Proc. IEEE International
beamforming with opportunistic scheduling,” IEEE Signal Process. Conference on Acoustics, Speech and Signal Processing (ICASSP),
Lett., vol. 20, no. 2, pp. 141–144, Feb. 2013. Vancouver, BC, Canada, May 2013, pp. 2906–2910.
[146] Y. Liu and A. P. Petropulu, “Destination assisted cooperative jamming [169] O. Ozel, K. Tutuncuoglu, J. Yang, S. Ulukus, and A. Yener, “Transmis-
for wireless physical-layer security,” IEEE Trans. Inf. Forensics Secu- sion with energy harvesting nodes in fading wireless channels: Optimal
rity, vol. 8, no. 4, pp. 682–694, Apr. 2013. policies,” IEEE J. Sel. Areas Commun., vol. 29, no. 8, pp. 1732–1743,
[147] J. Yang, I. M. Kim, and I. K. Dong, “Joint design of optimal cooperative Sep. 2011.
jamming and power allocation for linear precoding,” IEEE Trans. [170] J. Lei, Z. Han, M. A. Vazquez-Castro, and A. Hjorungnes, “Secure
Commun., vol. 62, no. 9, pp. 3285–3298, Sep. 2014. satellite communication systems design with individual secrecy rate
constraints,” IEEE Trans. Inf. Forensics Security, vol. 6, no. 3, pp.
[148] H.-M. Wang, F. Liu, and M. Yang, “Joint cooperative beamforming,
661–671, Sep. 2011.
jamming, and power allocation to secure AF relay systems,” IEEE
Trans. Veh. Technol., vol. 64, no. 10, pp. 4893–4898, Oct. 2015. [171] J. Li, A. P. Petropulu, and S. Weber, “Transmit power minimization
under secrecy capacity constraint in cooperative wireless commu-
[149] H. H. Kha, H. D. Tuan, and H. H. Nguyen, “Joint optimization of
nications,” in Proc. IEEE/SP 15th Workshop on Statistical Signal
source power allocation and cooperative beamforming for SC-FDMA
Processing, Cardiff, UK, Aug. 2009, pp. 217–220.
multi-user multi-relay networks,” IEEE Trans. Commun., vol. 61, no. 6,
[172] D. Wang, B. Bai, W. Chen, and Z. Han, “Secure green communication
pp. 2248–2259, Jun. 2013.
via untrusted two-way relaying: A physical layer approach,” IEEE
[150] H.-M. Wang, F. Liu, and X.-G. Xia, “Joint source-relay precoding
Trans. Commun., vol. 64, no. 5, pp. 1861–1874, May 2016.
and power allocation for secure amplify-and-forward MIMO relay
[173] X. Chen and L. Lei, “Energy-efficient optimization for physical layer
networks,” IEEE Trans. Inf. Forensics Security, vol. 9, no. 8, pp. 1240–
security in multi-antenna downlink networks with QoS guarantee,”
1250, Aug. 2014.
IEEE Commun. Lett., vol. 17, no. 4, pp. 637–640, Apr. 2013.
[151] C. Wang, H. M. Wang, D. W. K. Ng, and X. G. Xia, “Joint beamform-
[174] M. C. Gursoy, “Secure communication in the Low-SNR regime,” IEEE
ing and power allocation for secrecy in peer-to-peer relay networks,”
Trans. Commun., vol. 60, no. 4, pp. 1114–1123, Apr. 2012.
IEEE Trans. Wireless Commun., vol. 14, no. 6, pp. 3280–3293, Jun.
[175] C. Comaniciu and H. V. Poor, “On energy-secrecy tradeoffs for
2015.
Gaussian wire-tap channels,” IEEE Trans. Inf. Forensics Security,
[152] M. Bloch and J. Barros, Physical-Layer Security: From Information vol. 8, no. 8, pp. 314–323, Feb. 2013.
Theory to Security Engineering. Cambridge, U.K.: Cambridge Uni-
[176] R. Mahapatra, Y. Nijsure, G. Kaddoum, N. U. Hassan, and C. Yuen,
versity Press, 2011.
“Energy efficiency tradeoff mechanism towards wireless green com-
[153] Y. Liang, H. V. Poor, and S. Shamai, “Secure communication over munication: A survey,” IEEE Commun. Surveys Tuts., vol. 18, no. 1,
fading channels,” IEEE Trans. Inf. Theory, vol. 54, no. 6, pp. 2470– pp. 686–705, First Quarter 2016.
2492, Jun. 2008. [177] D. Feng, C. Jiang, G. Lim, L. J. Cimini, G. Feng, and G. Ye Li, “A
[154] J. Li, A. P. Petropulu, and S. Weber, “On cooperative relaying schemes survey of energy-efficient wireless communications,” IEEE Commun.
for wireless physical layer security,” IEEE Trans. Signal Process., Surveys Tuts., vol. 15, no. 1, pp. 167–178, First Quarter 2013.
vol. 59, no. 10, pp. 4985–4997, Oct. 2011. [178] D. W. K. Ng, E. S. Lo, and R. Schober, “Energy-efficient resource allo-
[155] M. Bloch, J. Barros, M. R. D. Rodrigues, and S. W. Mclaughlin, cation in multi-cell OFDMA systems with limited backhaul capacity,”
“Wireless information-theoretic security,” IEEE Trans. Inf. Theory, IEEE Trans. Wireless Commun., vol. 11, no. 10, pp. 3618–3631, Oct.
vol. 54, no. 6, pp. 2515–2534, Jun. 2008. 2012.
[156] S. Gerbracht, C. Scheunert, and E. A. Jorswieck, “Secrecy outage [179] E. Guvenkaya, J. M. Hamamreh, and H. Arslan, “On physical-layer
in MISO systems with partial channel information,” IEEE Trans. Inf. concepts and metrics in secure signal transmission,” Physical Commu-
Forensics Security, vol. 7, no. 2, pp. 704–716, Apr. 2012. nication, vol. 25, pp. 14–25, 2017.
[157] J. Barros and M. R. D. Rodrigues, “Secrecy capacity of wireless [180] M. Baldi, M. Bianchi, and F. Chiaraluce, “Coding with scrambling,
channels,” in Proc. IEEE International Symposium on Information concatenation, and HARQ for the AWGN wire-tap channel: A security
Theory (ISIT), Seattle, WA, Jul. 2006, pp. 356–360. gap analysis,” IEEE Trans. Inf. Forensics Security, vol. 7, no. 3, pp.
[158] S. Bashar, Z. Ding, and Y. G. Li, “On secrecy of codebook-based 883–894, Jun. 2012.
transmission beamforming under receriver limited feedback,” IEEE [181] J. M. Hamamreh, E. Basar, and H. Arslan, “OFDM-subcarrier index
Trans. Wireless Commun., vol. 10, no. 4, pp. 1212–1223, Apr. 2011. selection for enhancing security and reliability of 5G URLLC services,”
[159] J. Xiong, K.-K. Wong, D. Ma, and J. Wei, “A closed-form power IEEE Access, vol. 5, pp. 25 863–25 875, Dec 2017.
allocation for minimizing secrecy outage probability for MISO wiretap [182] J. M. Hamamreh, H. M. Furqan, Z. Ali, and G. A. S. Sidhu, “Enhancing
channels via masked beamforming,” IEEE Commun. Lett., vol. 16, the security performance of OSTBC using pre-equalicodization,” in
no. 9, pp. 1496–1499, Sep. 2012. Proc. International Conference on Frontiers of Information Technology
[160] M. Yuksel and E. Erkip, “Diversity-multiplexing tradeoff for the (FIT), Islamabad, Pakistan, Dec. 2017, pp. 294–298.
multiple-antenna wire-tap channel,” IEEE Trans. Wireless Commun., [183] M. Zhang, Y. Liu, and R. Zhang, “Artificial noise aided secrecy
vol. 10, no. 3, pp. 762–771, Mar. 2011. information and power transfer in OFDMA systems,” IEEE Trans.
[161] X. Zhou, M. R. Mckay, B. Maham, and A. Hjørungnes, “Rethinking the Wireless Commun., vol. 15, no. 4, pp. 3085–3096, Apr. 2016.
secrecy outage formulation: A secure transmission design perspective,” [184] N. Mokari, S. Parsaeefard, H. Saeedi, P. Azmi, and E. Hossain, “Secure
IEEE Commun. Lett., vol. 15, no. 3, pp. 302–304, Mar. 2011. robust ergodic uplink resource allocation in relay-assisted cognitive
 32

radio networks,” IEEE Trans. Signal Process., vol. 63, no. 2, pp. 291– [206] X. Xu, W. Yang, Y. Cai, and S. Jin, “On the secure spectral-energy
304, Jan. 2015. efficiency tradeoff in random cognitive radio networks,” IEEE J. Sel.
[185] D.P.Palomar, “A tutorial on decomposition methods for network utility Areas Commun., vol. 34, no. 10, pp. 2706–2722, Oct. 2016.
maximization,” IEEE J. Sel. Areas Commun., vol. 24, no. 8, pp. 1439– [207] Y. Kwon, H. Suh, J. Oh, and T. Hwang, “Energy efficient communi-
1451, Aug. 2006. cation for secure D2D underlaid cellular networks,” IEEE Trans. Veh.
[186] X. Wang, Y. Chen, L. Cai, and J. Pan, “Minimizing secrecy outage Technol., vol. 66, no. 10, pp. 9110–9123, Oct. 2017.
probability in multiuser wireless systems with stochastic traffic,” IEEE [208] W. Dinkelbach, “On nonlinear fractional programming,” Management
Trans. Veh. Technol., vol. 66, no. 7, pp. 6449–6460, Jul. 2017. Science, vol. 13, no. 7, pp. 492–498, Mar. 1967.
[187] T. X. Zheng, H. M. Wang, J. Yuan, D. Towsley, and M. H. Lee, “Multi- [209] S. Schaible, “Fractional programming II: on Dinkelbach’s algorithm,”
antenna transmission with artificial noise against randomly distributed Management Science, vol. 22, no. 8, pp. 868–873, Apr. 1976.
eavesdroppers,” IEEE Trans. Commun., vol. 63, no. 11, pp. 4347–4362, [210] T. F. Coleman and A. R. Conn, “Nonlinear programming via an exact
Nov. 2015. penalty function: Asymptotic analysis,” Mathematical Programming,
[188] H. Xing, L. Liu, and R. Zhang, “Secrecy wireless information and vol. 24, no. 1, pp. 137–161, 1982.
power transfer in fading wiretap channel,” IEEE Trans. Veh. Technol., [211] M. S. Bazaraa, H. D. Sherali, and C. M. Shetty, Nonlinear Program-
vol. 65, no. 1, pp. 180–190, Jan. 2016. ming: Theory and Algorithms. Hoboken, NJ, USA: Wiley, 2013.
[189] B. He and X. Zhou, “On the placement of RF energy harvesting node in [212] J. Gorski, F. Pfeuffer, and K. Klamroth, “Biconvex sets and optimiza-
wireless networks with secrecy considerations,” in Proc. IEEE Global tion with biconvex functions: A survey and extensions,” Math. Methods
Communications Conference (GLOBECOM) Workshops, Austin, TX, Oper., vol. 66, no. 3, pp. 373–407, Jun. 2007.
USA, Dec. 2014, pp. 1355–1360. [213] U. Niesen, D. Shah, and G. W. Wornell, “Adaptive alternating min-
[190] H. Yu, S. Guo, and Y. Yang, “An optimization framework of target imization algorithms,” IEEE Trans. Inf. Theory, vol. 55, no. 3, pp.
secrecy rate and power allocation for SWIPT system,” in Proc. IEEE 1423–1429, Mar. 2009.
Global Communications Conference (GLOBECOM), Washington, DC, [214] T. P. Dinh and H. A. L. Thi, “Recent advances in DC programming
USA, Dec. 2016, pp. 1–6. and DCA,” Transactions on Computational Intelligence XIII, pp. 1–37,
[191] J. Chen, X. Chen, and W. Gerstacker, “Optimal power allocation for 2014.
a massive MIMO relay aided secure communication,” in Proc. IEEE [215] L. T. H. An, M. T. Belghiti, and P. D. Tao, “A new efficient algorithm
Global Communications Conference (GLOBECOM), San Diego, CA, based on DC programming and DCA for clustering,” Journal of Global
USA, Feb. 2015, pp. 1–6. Optimization, vol. 37, no. 4, pp. 593–608, Aug. 2007.
[192] B. Akgun, O. O. Koyluoglu, and M. Krunz, “Exploiting full-duplex [216] A. O. Hero Iii, “Secure space-time communication,” IEEE Trans. Inf.
receivers for achieving secret communications in multiuser MISO Theory, vol. 49, no. 12, pp. 3235–3249, Dec. 2003.
networks,” IEEE Trans. Commun., vol. 65, no. 2, pp. 956–968, Feb. [217] S. Gong, C. Xing, S. Chen, and Z. Fei, “Secure communications for
2017. dual-polarized MIMO systems,” IEEE Trans. Signal Process., vol. 65,
[193] B. He, A. Liu, N. Yang, and V. K. N. Lau, “On the design of secure no. 16, pp. 4177–4192, Mar. 2017.
non-orthogonal multiple access systems,” IEEE J. Sel. Areas Commun.,
[218] Z. Q. Luo, W. K. Ma, M. C. So, Y. Ye, and S. Zhang, “Semidefinite
vol. 35, no. 10, pp. 2196–2206, Otc. 2017.
relaxation of quadratic optimization problems,” IEEE Signal Process.
[194] Y. Wu, K. Guo, J. Huang, and X. S. Shen, “Secrecy-based energy-
Mag., vol. 27, no. 3, pp. 20–34, May 2010.
efficient data offloading via dual connectivity over unlicensed spec-
[219] W. Ai, Y. Huang, and S. Zhang, “New results on hermitian matrix
trums,” IEEE J. Sel. Areas Commun., vol. 34, no. 12, pp. 3252–3270,
rank-one decomposition,” Mathematical Programming, vol. 128, no.
Dec. 2016.
1-2, pp. 253–283, 2011.
[195] M. El-Halabi, T. Liu, and C. N. Georghiades, “Secrecy capacity per
unit cost,” IEEE J. Sel. Areas Commun., vol. 31, no. 9, pp. 1909–1920, [220] L. Liu, R. Zhang, and K. C. Chua, “Secrecy wireless information and
Sep. 2013. power transfer with MISO beamforming,” IEEE Trans. Signal Process.,
[196] M. M. Butt, E. A. Jorswieck, and B. Ottersten, “Maximizing energy vol. 62, no. 7, pp. 1850–1863, Apr. 2014.
efficiency in multiple access channels by exploiting packet dropping [221] M. R. A. Khandaker and K. K. Wong, “Masked beamforming in
and transmitter buffering,” IEEE Trans. Wireless Commun., vol. 14, the presence of energy-harvesting eavesdroppers,” IEEE Trans. Inf.
no. 8, pp. 4129–4141, Aug. 2015. Forensics Security, vol. 10, no. 1, pp. 40–54, Jan. 2015.
[197] A. Zappone, P. H. Lin, and E. A. Jorswieck, “Energy-efficient secure [222] Y. Huang and D. P. Palomar, “Rank-constrained separable semidefinite
communications in MISO-SE systems,” in Proc. 48th Asilomar Con- programming with applications to optimal beamforming,” IEEE Trans.
ference on Signals, Systems and Computers, Pacific Grove, CA, Nov. Signal Process., vol. 58, no. 2, pp. 664–678, Feb. 2010.
2014, pp. 1001–1005. [223] S. Yan and R. Malaney, “Location-based beamforming for enhancing
[198] ——, “Energy efficiency of confidential multi-antenna systems with secrecy in Rician wiretap channels,” IEEE Trans. Wireless Commun.,
artificial noise and statistical CSI,” IEEE J. Sel. Topics Signal Process., vol. 15, no. 4, pp. 2780–2791, Apr. 2016.
vol. 10, no. 8, pp. 1462–1477, Dec. 2016. [224] C. Liu and R. Malaney, “Location-based beamforming and physical
[199] ——, “Optimal energy-efficient design of confidential multiple-antenna layer security in Rician wiretap channels,” IEEE Trans. Wireless
systems,” IEEE Trans. Inf. Forensics Security, vol. 13, no. 1, pp. 237– Commun., vol. 15, no. 11, pp. 7847–7857, Nov. 2016.
252, Jan. 2018. [225] Y. R. Ramadan, H. Minn, and A. S. Ibrahim, “Hybrid analog-digital
[200] H. Q. Ta and S. W. Kim, “Adapting rate and power for maximizing precoding design for secrecy mmWave MISO-OFDM systems,” IEEE
secrecy energy efficiency,” IEEE Commun. Lett., vol. 21, no. 9, pp. Trans. Wireless Commun., vol. 65, no. 11, pp. 5009–5026, Nov. 2017.
2049–2052, Sep. 2017. [226] J. M. Kang, J. Yang, J. Ha, and I. M. Kim, “Joint design of optimal
[201] J. Farhat, G. Brante, R. D. Souza, and J. L. Rebelatto, “Energy precoding and cooperative jamming for multi-user secure broadcast
efficiency of repetition coding and parallel coding relaying under partial systems,” IEEE Trans. Veh. Technol., vol. 66, no. 11, pp. 10 551–10 556,
secrecy regime,” IEEE Access, vol. 4, pp. 7275–7288, Nov. 2016. Nov. 2017.
[202] F. Gabry, A. Zappone, R. Thobaben, E. A. Jorswieck, and M. Skoglund, [227] F. Zhu, F. Gao, T. Zhang, K. Sun, and M. Yao, “Physical-layer security
“Energy efficiency analysis of cooperative jamming in cognitive radio for full duplex communications with self-interference mitigation,”
networks with secrecy constraints,” IEEE Wireless Commun. Lett., IEEE Trans. Wireless Commun., vol. 15, no. 1, pp. 329–340, Jan. 2016.
vol. 4, no. 4, pp. 437–440, Aug. 2015. [228] F. Zhu and M. Yao, “Improving physical-layer security for CRNs using
[203] X. Xu, Y. Cai, W. Yang, and W. Yang, “Energy-efficient optimization SINR-based cooperative beamforming,” IEEE Trans. Veh. Technol.,
for physical layer security in large-scale random CRNs,” in Proc. vol. 65, no. 3, pp. 1835–1841, Mar. 2016.
International Conference on Wireless Communications and Signal [229] Z. Zhu, Z. Chu, Z. Wang, and I. Lee, “Outage constrained robust
Processing, Nanjing, China, Oct. 2015. beamforming for secure broadcasting systems with energy harvesting,”
[204] J. Ouyang, W. P. Zhu, D. Massicotte, and M. Lin, “Energy efficient IEEE Trans. Wireless Commun., vol. 15, no. 11, pp. 7610–7620, Nov.
optimization for physical layer security in cognitive relay networks,” in 2016.
Proc. IEEE International Conference on Communications (ICC), Kuala [230] H. Zhang, Y. Huang, C. Li, and L. Yang, “Secure beamforming design
Lumpur, Malaysia, May. 2016, pp. 1–6. for SWIPT in MISO broadcast channel with confidential messages
[205] R. Zhang, C. Comaniciu, and H. V. Poor, “Outage capacity and partial and external eavesdroppers,” IEEE Trans. Wireless Commun., vol. 15,
secrecy for energy efficient physical layer security in Gaussian fading no. 11, pp. 7807–7819, Nov. 2016.
channels,” in Proc. 16th International Symposium on Wireless Personal [231] Z. Zhu, Z. Chu, N. Wang, S. Huang, Z. Wang, and I. Lee, “Beam-
Multimedia Communications (WPMC), Atlantic City, NJ, Jun. 2013. forming and power splitting designs for AN-aided secure multi-user
 33

MIMO SWIPT systems,” IEEE Trans. Inf. Forensics Security, vol. 12, [252] J. Mo, M. Tao, and Y. Liu, “Relay placement for physical layer security:
no. 12, pp. 2861–2874, Dec. 2017. A secure connection perspective,” IEEE Commun. Lett., vol. 16, no. 6,
[232] D. W. K. Ng and R. Schober, “Resource allocation for secure com- pp. 878–881, Jun. 2012.
munication in systems with wireless information and power transfer,” [253] C. Wang and H.-M. Wang, “Robust joint beamforming and jamming
in Proc. IEEE Global Communications Conference (GLOBECOM) for secure AF networks: Low-complexity design,” IEEE Trans. Veh.
Workshops, Atlanta, GA, USA, Dec. 2013, pp. 1251–1257. Technol., vol. 64, no. 5, pp. 2192–2198, May 2014.
[233] S. Leng, D. W. K. Ng, and R. Schober, “Power efficient and secure [254] M. Lin, J. Ge, Y. Yang, and Y. Ji, “Joint cooperative beamforming and
multiuser communication systems with wireless information and power artificial noise design for secrecy sum rate maximization in two-way
transfer,” in Proc. IEEE International Conference on Communications AF relay networks,” IEEE Commun. Lett., vol. 18, no. 2, pp. 380–383,
(ICC) Workshops, Sydney, NSW, Australia, Jun. 2014, pp. 800–806. Feb. 2014.
[234] D. W. K. Ng, E. S. Lo, and R. Schober, “Robust beamforming for [255] Z. Ding, M. Peng, and H. Chen, “A general relaying transmission
secure communication in systems with wireless information and power protocol for MIMO secrecy communications,” IEEE Trans. Commun.,
transfer,” IEEE Trans. Wireless Commun., vol. 13, no. 8, pp. 4599– vol. 60, no. 11, pp. 3461–3471, Nov. 2012.
4615, Aug. 2014. [256] Y. Huang, F. S. Al-Qahtani, T. Q. Duong, and J. Wang, “Secure
[235] D. W. K. Ng, L. Xiang, and R. Schober, “Multi-objective beamforming transmission in MIMO wiretap channels using general-order transmit
for secure communication in systems with wireless information and antenna selection with outdated CSI,” IEEE Trans. Commun., vol. 63,
power transfer,” in Proc. IEEE 24th Annual International Symposium no. 8, pp. 2959–2971, Aug. 2015.
on Personal, Indoor, and Mobile Radio Communications (PIMRC), [257] R. Zhao, H. Lin, Y. C. He, D. H. Chen, Y. Huang, and L. Yang,
London, UK, Sep. 2013, pp. 7–12. “Secrecy performance of transmit antenna selection for MIMO relay
systems with outdated CSI,” IEEE Trans. Commun., vol. 66, no. 2, pp.
[236] D. W. K. Ng and R. Schober, “Secure and green SWIPT in distributed
546–559, 2018.
antenna networks with limited backhaul capacity,” IEEE Trans. Wire-
[258] H. Lei, J. Zhang, K. H. Park, P. Xu, I. S. Ansari, G. Pan, B. Alomair,
less Commun., vol. 14, no. 9, pp. 5082–5097, Sep. 2015.
and M. S. Alouini, “On secure noma systems with transmit antenna
[237] O. Tervo, H. Pennanen, D. Christopoulos, S. Chatzinotas, and B. Otter- selection schemes,” IEEE Access, vol. 5, no. 99, pp. 17 450–17 464,
sten, “Distributed optimization for coordinated beamforming in multi- Aug 2017.
cell multigroup multicast systems: Power minimization and SINR [259] G. Chen, Y. Gong, P. Xiao, and J. A. Chambers, “Dual antenna selec-
balancing,” IEEE Trans. Signal Process., vol. 66, no. 1, pp. 171–185, tion in secure cognitive radio networks,” IEEE Trans. Veh. Technol.,
Jun. 2018. vol. 65, no. 10, pp. 7993–8002, Oct. 2015.
[238] H. Zhang, P. Sun, C. Li, Y. Huang, and L. Yang, “Cooperative [260] Z. Ding, Z. Ma, and P. Fan, “Asymptotic studies for the impact of
precoding for wireless energy transfer and secure cognitive radio antenna selection on secure two-way relaying communications with
coexistence systems,” IEEE Signal Process. Lett., vol. 24, no. 5, pp. artificial noise,” IEEE Trans. Wireless Commun., vol. 13, no. 4, pp.
540–544, May. 2017. 2189–2203, Apr. 2014.
[239] D. W. K. Ng, E. S. Lo, and R. Schober, “Multi-objective resource [261] Z. Lin, Y. Cai, W. Yang, and L. Wang, “Robust secure switching
allocation for secure communication in cognitive radio networks with transmission in multi-antenna relaying systems: cooperative jamming
wireless information and power transfer,” IEEE Trans. Veh. Technol., or decode-and-forward beamforming,” IET Communications, vol. 10,
vol. 65, no. 5, pp. 3166–3184, May 2016. no. 13, pp. 1673–1681, Sep. 2016.
[240] A. Beck, A. Ben-Tal, and L. Tetruashvili, “A sequential parametric [262] Q. T. Vien, T. A. Le, and T. Q. Duong, “Opportunistic secure transmis-
convex approximation method with applications to nonconvex truss sion for wireless relay networks with modify-and-forward protocol,” in
topology design problems,” J. Global Optim., vol. 47, no. 1, pp. 29– Proc. IEEE International Conference on Communications (ICC), Paris,
51, May 2010. France, May 2017, pp. 1–6.
[241] A. Canelas, M. Carrasco, and J. Lpez, “Application of the sequential [263] T. M. Hoang, T. Q. Duong, N. S. Vo, and C. Kundu, “Physical layer
parametric convex approximation method to the design of robust security in cooperative energy harvesting networks with a friendly
trusses,” J. Global Optim., vol. 68, no. 1, pp. 1–19, May 2017. jammer,” IEEE Commun. Lett., vol. 6, no. 2, pp. 174–177, Apr. 2017.
[242] Y. Nesterov, “Semidefinite relaxation and nonconvex quadratic opti- [264] J. Moon, H. Lee, C. Song, and I. Lee, “Secrecy outage minimization
mization,” Optimization Methods & Software, vol. 9, no. 1, pp. 141– for wireless powered communication networks with an energy har-
160, 1998. vesting jammer,” in Proc. IEEE Global Communications Conference
[243] N. T. Nghia, H. D. Tuan, T. Q. Duong, and H. V. Poor, “MIMO (GLOBECOM), Washington, DC, USA, Dec. 2016, pp. 1–5.
beamforming for secure and energy-efficient wireless communication,” [265] ——, “Secrecy performance optimization for wireless powered com-
IEEE Signal Process. Lett., vol. 24, no. 2, pp. 236–239, Feb. 2017. munication networks with an energy harvesting jammer,” IEEE Trans.
[244] W. Mei, Z. Chen, and J. Fang, “Robust energy-efficient transmit design Commun., vol. 65, no. 2, pp. 764–774, Feb. 2017.
for MISOME wiretap channels,” in Proc. IEEE Global Conference [266] A. Mabrouk, A. E. Shafie, K. Tourki, and N. Al-Dhahir, “An-aided
on Signal and Information Processing (GlobalSIP), Washington, DC, relay-selection scheme for securing untrusted RF-EH relay systems,”
USA, Dec. 2016, pp. 981–985. IEEE Trans. Green Commun. Networking, vol. 1, no. 4, pp. 481–493,
[245] J. Ouyang, M. Lin, W. P. Zhu, D. Massicotte, and A. L. Swindlehurst, Dec. 2017.
“Energy efficient beamforming for secure communication in cognitive [267] J. Farhat, G. Brante, and R. D. Souza, “On the secure energy efficiency
radio networks,” in Proc. IEEE International Conference on Acoustics, of TAS/MRC with relaying and jamming strategies,” IEEE Signal
Speech and Signal Processing (ICASSP), Shanghai, China, Mar. 2016, Process. Lett., vol. 24, no. 8, pp. 1228–1232, Aug. 2017.
pp. 3496–3500. [268] L. Wang, L. L. Yang, V. C. M. Schober, and M. Song, “Adaptive
cooperation schemes for energy efficient physical layer security,” in
[246] W. Mei, Z. Chen, and J. Fang, “Artificial noise aided energy efficiency
Proc. IEEE INFOCOM Workshops, Toronto, Canada, Apr. 2014, pp.
optimization in MIMOME system with SWIPT,” IEEE Commun. Lett.,
159–160.
vol. 21, no. 8, pp. 1795–1798, Aug. 2017.
[269] T. X. Zheng, H. M. Wang, J. Yuan, Z. Han, and M. H. Lee, “Physical
[247] H. Alves, M. D. C. Tome, P. H. J. Nardelli, C. H. M. D. Lima, and layer security in wireless ad hoc networks under a hybrid full-/half-
M. Latva-Aho, “Enhanced transmit antenna selection scheme for secure duplex receiver deployment strategy,” IEEE Trans. Wireless Commun.,
throughput maximization without CSI at the transmitter,” IEEE Access, vol. 16, no. 6, pp. 3827–3839, Jun. 2017.
vol. 4, pp. 4861–4873, Aug. 2016. [270] L. Wang, K. K. Wong, M. Elkashlan, A. Nallanathan, and S. Lam-
[248] R. Liu and W. Trappe, Securing Wireless Communications at the botharan, “Secrecy and energy efficiency in massive MIMO aided
Physical Layer. New York, USA: Springer-Verlag, 2010. heterogeneous C-RAN: A new look at interference,” IEEE J. Sel. Topics
[249] L. Lai and H. El Gamal, “The relay-eavesdropper channel: Cooperation Signal Process., vol. 10, no. 8, pp. 1375–1389, Aug. 2016.
for secrecy,” IEEE Trans. Inf. Theory, vol. 54, no. 9, pp. 4005–4019, [271] J. Chen, X. Chen, T. Liu, and L. Lei, “Toward green and secure
Sep. 2008. communications over massive MIMO relay networks: Joint source and
[250] S. Vishwakarma and A. Chockalingam, “Decode-and-forward relay relay power allocation,” IEEE Access, vol. 5, pp. 869–880, Jan. 2017.
beamforming for secrecy with finite-alphabet input,” IEEE Commun. [272] Y. Kwon, X. Wang, and T. Hwang, “A game with randomly distributed
Lett., vol. 17, no. 5, pp. 912–915, May 2013. eavesdroppers in wireless ad hoc networks: A secrecy EE perspective,”
[251] T. X. Zheng, H.-M. Wang, F. Liu, and M. H. Lee, “Outage constrained IEEE Trans. Veh. Technol., vol. 66, no. 11, pp. 9916–9930, Nov. 2017.
secrecy throughput maximization for DF relay networks,” IEEE Trans. [273] J. Li, A. P. Petropulu, and H. V. Poor, “Cooperative transmission
Commun., vol. 63, no. 5, pp. 1741–1755, May. 2015. for relay networks based on second-order statistics of channel state
 34

information,” IEEE Trans. Signal Process., vol. 59, no. 3, pp. 1280–
1291, Mar. 2011.
[274] A. Goldsmith, S. A. Jafar, N. Jindal, and S. Vishwanath, “Capacity
limits of MIMO channels,” IEEE J. Sel. Areas Commun., vol. 21, no. 5,
pp. 684–702, Jun. 2003.
[275] S. Ma, M. Hong, E. Song, X. Wang, and D. Sun, “Outage constrained
robust secure transmission for MISO wiretap channels,” IEEE Trans.
Wireless Commun., vol. 13, no. 10, pp. 5558–5570, Oct. 2013.
[276] X. Chen, L. Lei, H. Zhang, and C. Yuen, “Large-scale MIMO relay-
ing techniques for physical layer security: AF or DF?” IEEE Trans.
Wireless Commun., vol. 14, no. 9, pp. 5135–5146, Sep. 2015.
[277] G. Brante, H. Alves, R. D. Souza, and M. Latva-aho, “Secrecy analysis
of transmit antenna selection cooperative schemes with no channel state
information at the transmitter,” IEEE Trans. Commun., vol. 63, no. 4,
pp. 1330–1342, Apr. 2015.
[278] P. Kyritsi, D. C. Cox, R. A. Valenzuela, and P. W. Wolniansky,
“Correlation analysis based on MIMO channel measurements in an
indoor environment,” IEEE J. Sel. Areas Commun., vol. 21, no. 5, pp.
713–720, Jun. 2003.
[279] X. He, H. Dai, Y. Huang, D. Wang, W. Shen, and P. Ning, “The
security of link signature: A view from channel models,” in Proc. IEEE
Conference on Communications and Network Security, San Francisco,
CA, USA, Oct. 2014, pp. 103–108.
[280] Y. Liang, H. V. Poor, and L. Ying, “Secure communications over
wireless broadcast networks: Stability and utility maximization,” IEEE
Trans. Inf. Forensics Security, vol. 6, no. 3, pp. 682–692, Sep. 2011.
[281] X. Tang, P. Ren, Y. Wang, and Z. Han, “Combating full-duplex
active eavesdropper: A hierarchical game perspective,” IEEE Trans.
Commun., vol. 65, no. 3, pp. 1379–1395, Mar. 2017.
[282] M. R. Abedi, N. Mokari, H. Saeedi, and H. Yanikomeroglu, “Robust
resource allocation to enhance physical layer security in system with
full-duplex receivers: A ctive adversary,” IEEE Trans. Wireless Com-
mun., vol. 16, no. 2, pp. 885–899, Feb. 2017.
[283] E. Boshkovska, D. W. K. Ng, L. Dai, and R. Schober, “Power-efficient
and secure WPCNs with hardware impairments and non-linear EH
circuit,” IEEE Trans. Commun., vol. 66, no. 6, pp. 2642–2657, Jun.
2018.
[284] E. Bjornson, M. Matthaiou, and M. Debbah, “A new look at dual-hop
relaying: Performance limits with hardware impairments,” IEEE Trans.
Commun., vol. 61, no. 11, pp. 4512–4525, Nov. 2013.
[285] U. Gustavsson, C. Sanchez-Perez, T. Eriksson, and F. Athley, “On
the impact of hardware impairments on massive MIMO,” in Proc.
IEEE Global Communications Conference (GLOBECOM) Workshops,
Austin, TX, USA, Dec. 2014, pp. 294–300.