 Research
 Open Access
Artificial noiseassisted physical layer security in D2Denabled cellular networks
 Yajun Chen^{1}Email author,
 Xinsheng Ji^{1, 2, 3},
 Kaizhi Huang^{1},
 Jing Yang^{1},
 Xin Hu^{1} and
 Yunjia Xu^{1}
https://doi.org/10.1186/s1363801709691
© The Author(s) 2017
 Received: 27 July 2017
 Accepted: 24 October 2017
 Published: 2 November 2017
The Correction to this article has been published in EURASIP Journal on Wireless Communications and Networking 2017 2017:199
Abstract
Devicetodevice (D2D) communication has been deemed as a promising technology in the next generation 5G wireless communication. Due to the openness nature of the transmission medium, secure transmission is also a critical issue in the D2Denabled cellular network as well as other wireless systems. In this paper, we investigate secure communication for the cellular downlink in this hybrid network. We consider a case in which each base station has no channel state information (CSI) from D2D transmitters which are generally deployed in the cell edge. To guarantee the secure communication of the cellular link, each base station employs the artificial noise assisted transmission strategy. Firstly, we derive the closeform expression and asymptotic expression of the secrecy outage probability of the cellular link in different scenarios: (I) eavesdroppers having no multiuser decedability; (II) eavesdroppers having the multiuser decedability. Then, we comprehensively discuss the impacts of some main system parameters on the performance to provide some system design guidances. To characterize the reliable communication of the typical D2D link, the closeform expression and asymptotic expression of the connection outage probability are, respectively, derived and some comprehensive analysis are presented. Finally, simulation results are provided to validate the effectiveness of theoretical analysis.
Keywords
 Devicetodevice (D2D) communication
 Physical layer security
 Artificial noise
 Secrecy outage probability
 Connection outage probability
1 Introduction
To meet the explosive demand of proximity services, devicetodevice (D2D) communication has been regarded as an ideal candidate technology for the next generation 5G wireless communication. D2D communication allows proximity user equipments to deliver their own messages over the direct link established between them without the base station relaying messages, which has the promise of many types of advantages: superior spectrum efficiency, increasing quality of service (QoS) of edge users and network capacity. Accordingly, D2D communication underlaying a cellular network has attracted remarkable attention both in the world of academia [1–3] and industry [4, 5] in recent years.
Due to the openness nature of the transmission medium, secure transmission is identified as a critical challenge facing the D2Denabled cellular network as well as other wireless systems. As a remedy of the traditional security mechanism, the concept of physical layer security (PHYsecurity) has been proposed recently to achieve secure communication for wireless systems by exploiting the characteristics of wireless channels.
1.1 Related works
Recently, PHYsecurity in the D2Denabled cellular network has sparked of wide interests and achieved fruitful research works in many different scenarios. To the best of our knowledge, most of works designed different resourcescheduling schemes to guarantee secure communication for the cellular uplink, such as [6–11]. More specifically, the literature above employed the hybrid interference to improve secure performance for the cellular uplink.
For the cellular downlink, Liu et al. proposed a power transfer model and an information signal model to enable wireless energy harvesting and secure information transmission in largerscale cognitive cellular networks and comprehensively discussed wireless power transfer policies and secrecy performance in [12]. More particularly, the authors in [13] designed two optimal D2D linkscheduling schemes under different criteria when each base station has a single antenna. When a base station is equipped with multiantenna, the space redundancy could be exploited to enhance secure communication for the cellular link through some designed schemes, such as the secure beamforming and artificial noise assisted scheme. Specifically, Chu et al. investigated robust secrecy rate optimization problems for a multipleinput singleoutput (MISO) secrecy channel with multiple D2D communications, which was equivalently converted into the power minimization problem and the secrecy rate maximization problem to design the robust secure beamforming in [14]. The authors in [15] investigated a secure wireless powered D2D communication, in which a base station charged for D2D transmitters in wireless energy transfer phase and introduced the jamming service to interfere with the multiple eavesdroppers.
On the other hand, artificial noiseassisted scheme in the field of PHYsecurity is the most representative one among different schemes assuring the security of the wireless communication in MISO or multiinput multioutput (MIMO) scenario [16, 17]. The design idea of the artificial noise assisted scheme is that legitimate transmitters inject artificial noise into their transmission signals to confuse malicious eavesdroppers. Meanwhile, in order to guarantee the reliable communication of the legitimate user as much as possible, the artificial noise should be injected into the null space of the main channel (from source to destination). The authors in [18, 19] have expended the artificial noiseassisted scheme to the MISO D2Denabled cellular network. More particularly, they designed the corresponding artificial noiseassisted beamforming vector matrix under the assumption that the channel state information (CSI) from each D2D transmitter is perfectly known at each base station.
Nevertheless, they only considered one cellular user and one D2D pair within a cell, which only focused on the pointtopoint link and ignored the interference from other neighbor cells [18, 19]. On the other hand, due to the original purpose of the D2D communication, the D2D transmitter is generally deployed in the cell edge. Hence, in practical cases, CSI between each base station and each D2D transmitter is difficult to be perfectly known at each base station due to the channel estimation and quantization errors.
1.2 Motivation and contributions

In this hybrid network, we consider a case in which each base station has no CSI from each D2D transmitter generally deployed in the cell edge. To guarantee the secure communication for the cellular link, it is assumed that each base station employs the artificial noiseassisted transmission scheme. The closeform expression and asymptotic expression of secrecy outage probability of the typical cellular link are derived in the scenario in which eavesdroppers do not have the multiuser decodability. Based on the derived result of secrecy outage probability, we provide some comprehensive analysis on the secrecy performance of the cellular link.

Then, when eavesdroppers have the multiuser decodability, the closeform expression and asymptotic expression of secrecy outage probability of the typical cellular link are also, respectively, derived. Based on derived results of the secrecy outage probability in this case, some comprehensive analysis are also provided to guide the system design.

Finally, according to the design of the artificial noise assisted transmission scheme, the artificial noise is just injected into the null space of the cellular channel because each base station does not know CSI from each D2D transmitter. Hence, the artificial noise and the informationbearing signal could all degrade the reliable communication of the typical D2D link. In order to characterize the reliable performance, we analytically derive the closeform expression and the asymptotic expression of connection outage probability, and provide some comprehensive analysis.
1.3 Organization and notations
The reminder of this paper is organized as follows. In Section 2, we present the system model. In Section 3, the secrecy outage probability of the typical cellular link, the connection outage probability of the typical D2D link are respectively derived and some corresponding analysis are provided. Simulation results are presented in Section 4. Finally, we conclude this paper in Section 5.
Notations: Bold letters mean matrices (column vectors). We use \({{\mathcal C}{\mathcal N}}\left ({\mu,{N_{0}}} \right)\) to denote the circularly symmetric complex Gaussian with mean μ and covariance N _{0}. \(\mathbb {P}\left \{\bullet \right \}\) represents the probability of an input event and the notation \({\mathbb {E}}\{\bullet \}\) denotes the statistical expectation. exp(1) denotes the exponential distribution with unit mean. Gamma(N,λ) is Gamma distribution with parameters N and λ. In addition, ∥∙∥ denotes euclidean norm and (∙)^{ T } means the transpose of the input matrix. \({\kappa _{n}} \buildrel \Delta \over = {{\Gamma \left ({n  1 + \rho } \right)\Gamma \left ({1  \rho } \right)} \left /\right. {\Gamma \left ({n  1} \right)}}\) and Γ(x) is gamma function.
2 System model
2.1 Network model
Both the largescale fading and smallscale fading of wireless channels are considered in this paper. The standard path loss model is taken into account for the largescale fading, i.e., l(r _{ ij })=r _{ ij } ^{−α }, where r _{ ij } is the distance between the node i and the node j, and α>2 represents the fading coefficient. In addition, the smallscale fading imposes the independent quasistatic Rayleigh fading model, whose coefficient is constant for each transmission block.
2.2 Artificial noiseassisted transmission scheme and wiretap code
where s _{ i } is the informationbearing signal with \({\mathbb {E}\left [ {{{\left  {{s_{i}}} \right }^{2}}} \right ] = 1}\), \({{\mathbf {v}}_{i} \in \mathbb {C}^{\left ({M1}\right) \times {1}}}\) is an artificial noise vector with independent identically distributed (i.i.d.) entries \({v_{i,n}}\sim {{\mathcal {C}}{\mathcal {N}}}\left ({0,\frac {1}{{M  1}}} \right)\). p _{ I }=ϕ p is the allocated transmission power to the informationbearing signal and p _{ A }=(1−ϕ)p represents the allocated transmission power to generate the artificial noise at each base station to confuse malicious eavesdroppers, where p is the total transmission power. Thus, ϕ∈[0,1] represents the ratio of the total transmission power p allocated to transmit the informationbearing signal s _{ i }. In addition, h _{ i } means the wireless channel between each base station and the served cellular user. We consider a case in which it is assumed that CSI between each base station and D2D users is unknown at each base station because D2D users are generally deployed in the cell edge in practical cases. According to the design of the artificial noiseassisted transmission scheme, the beamforming vector w _{ i } for the served cellular user should satisfy \({{\mathbf {w}}_{i}} = {{{\mathbf {h}}_{i}^ +} \left /\right. {\left \ {{{\mathbf {h}}_{i}}} \right \}}\). \({{\mathbf {W}}_{i}} \in \mathbb {C}^{M \times \left ({M  1} \right)}\) is a weight matrix for the artificial noise, and the columns of \({\mathbf {W}} \buildrel \Delta \over = \left [ {{{\mathbf {w}}_{i}} {{\mathbf {W}}_{i}}} \right ]\) constitute an orthogonal basis.
To improve the secrecy performance of this hybrid network, all transmitters adopt the wiretap code scheme to encode the data before transmission. More specifically, it is assumed that the rate of the transmitted codeword and the rate of the confidential message are, respectively, denoted by R _{ b } and R _{ s }. The codeword rate R _{ b } is the actual transmission rate of the codewords, while the secrecy rate R _{ s } is the rate of the embedded message. The rate redundancy R _{ e }=R _{ b }−R _{ s } is intentionally added in order to provide secrecy against malicious eavesdroppers. More discussions on code construction can be found in [20].
2.3 Performance metrics
In this paper, we mainly focus on two performance metrics: secrecy outage probability and connection outage probability to respectively characterize the performance of the cellular link and D2D link in this hybrid network. Based on the analysis above, we next will give their definitions.
When the capacity of the channel from the legitimate transmitter to the corresponding receiver falls below the predefined target codeword rate R _{ b }, the receiver will not decode the transmission message correctly. We define the probability of this event as connection outage probability [21].
In addition, when the capacity of the most detrimental one among multiple eavesdroppers (i.e., the eavesdropper having the maximal capacity of the channel from the legitimate transmitter to multiple eavesdroppers) is above the predefined target rate redundancy R _{ e }, confidential messages for legitimate receivers will be decoded correctly and obtained by malicious eavesdroppers. We define the probability of this event as secrecy outage probability [21]^{1}.
where SINR and SINR_{ e } respectively, denote the received SINR at the legitimate receiver and the most detrimental eavesdropper. α and β are the target SINR thresholds for reliable and secure communication, respectively.
2.4 Cellular user association
where ς=3.5 for the nearest base station association scheme and δ=λ _{ c }/λ _{ b } represents the cell load. Note that when there are more than one cellular users to be served by a base station, the base station just chooses one cellular user to serve at each time slot through the timedivision multiple access (TDMA) scheme^{2}.
3 Outage probabilities analysis
In the section, we will conduct the performance analysis about the security of the typical cellular link and the reliability of the typical D2D link, respectively. Firstly, considering whether eavesdroppers have the multiuser decedability or not, we derive the closeform expression and asymptotic expression of the secrecy outage probability of the cellular link in two different scenarios in the noncolluding way. Then, for the typical D2D link, we consider its reliable communication and derive the closeform expression and asymptotic expression of the connection outage probability.
3.1 Secrecy outage probability of cellular links
In this subsection, we will conduct the secrecy performance analysis for the cellular link and derive the secrecy outage probability of the cellular link in two different scenarios depending on whether eavesdroppers have the multiuser decedability or not. Due to the property of PPP that its distribution will not be changed by shifting the coordinates, we firstly shift the coordinates to put the typical base station located at the origin.
Note that the rate redundancy is R _{ e } and \({\hat \gamma _{e}} = {2^{{R_{e}}}}  1\). (a) follows from the independent of different channel gains. (b) follows from the probability generating functional (PGFL) of PPP [24]: \({}\left [ {\prod \limits _{x \in \Phi } {f\left (x \right)}} \right ] = \exp \left (\lambda \int _{{R^{2}}} {\left ({1  f\left (x \right)} \right)dx} \right)\). \(\Phi ^{a}_{b}\) represents the active base station set.
On the other hand, since eavesdroppers generally work in a passive way, it is difficult for legitimate transmitters to know their abilities to overhear the confidential message of the cellular link. Hence, according to the different abilities of eavesdroppers to decode the transmission message, we consider the secrecy performance of the cellular link and derive the respective expressions of the secrecy outage probability in two different scenarios in the following subsection. Firstly, we will discuss the performance of the cellular link in the case in which eavesdroppers have no multiuser decedability.
3.1.1 Scenario I
where \(\frac {{{p_{A}}}}{{M  1}}{{\left \ {{\mathbf {g}}_{0e}^{T}{{\mathbf {W}}_{0}}} \right \}^{2}}\) represents the received interference induced by the injected artificial noise from the typical base station. \(I_{e{\backslash \left \{ 0 \right \}}}=\sum \limits _{{y_{i}} \in {\Phi _{d}}} {{p_{d}}{{\left  {{h_{i}}} \right }^{2}}{{\left \ {{y_{i}}  {x_{z}}} \right \}^{ \alpha }}}+\sum \limits _{{x_{i}} \in {\Phi ^{a}_{b}}\backslash \left \{ 0 \right \}} {\left ({{p_{I}}{{\left  {{\mathbf {g}}_{ie}^{T}{{\mathbf {w}}_{i}}} \right }^{2}}{+ }\frac {{{p_{A}}}}{{M  1}}{{\left \ {{\mathbf {g}}_{ie}^{T}{{\mathbf {W}}_{i}}} \right \}^{2}}} \right){{\left \ {{x_{i}}  {x_{z}}} \right \}^{ \alpha }}}\) represents the cumulative interference from legitimate transmitters (including both D2D transmitters located at y _{ i } and base stations located at x _{ i }, but except the typical base station located at the origin). For notational conciseness, we define \(I_{e,ce}=\sum \limits _{{x_{i}} \in {\Phi ^{a}_{b}}\backslash \left \{ 0 \right \}} {\left ({{p_{I}}{{\left  {{\mathbf {g}}_{ie}^{T}{{\mathbf {w}}_{i}}} \right }^{2}} + \frac {{{p_{A}}}}{{M  1}}{{\left \ {{\mathbf {g}}_{ie}^{T}{{\mathbf {W}}_{i}}} \right \}^{2}}} \right){{\left \ {{x_{i}}  {x_{z}}} \right \}^{ \alpha }}}\), which represents the cumulative interference from other base stations located at x _{ i } (except the typical base station) induced by both the informationbearing signal and the artificial noise. \( I_{e,d  e}=\sum \limits _{{y_{i}} \in {\Phi _{d}}} {{p_{d}}{{\left  {{h_{i}}} \right }^{2}}{{\left \ {{y_{i}}  {x_{z}}} \right \}^{ \alpha }}}\) represents the cumulative interference from all the D2D transmitters. N _{0} represents the covariance of the additive Gaussian noise at eavesdroppers. Based the analysis above, we can easily obtain I _{ e∖{0}}=I _{ e,c−e }+I _{ e,d−e }.
Theorem 1
Considering the case in which eavesdroppers have no multiuser decedability, the closeform expression of the secrecy outage probability of the cellular link can be given by:
where \(\rho = \frac {2}{\alpha }\) and \(s = \frac {{{{\hat \gamma }_{e}}{r^{\alpha } }}}{{{p_{I}}}}\). For notational conciseness, we define \(\upsilon = 2\pi {\lambda _{e}}{{\left ({1 + {{\hat \gamma }_{e}}\xi } \right)}^{1  M}}\) and \(\mu = { \left ({\pi {\lambda _{b}}{p_{a}}\omega p_{I}^{\rho } + \frac {{\pi {\lambda _{d}}p_{d}^{\rho } }}{{\sin c\rho }}} \right){s^{\rho } }}\), where ξ=(ϕ ^{−1}−1)/(M−1) and \(\frac {1}{{\sin c\rho }} = \frac {{\pi \rho }}{{\sin \pi \rho }}=\Gamma \left ({1 + \rho } \right)\Gamma \left ({1  \rho } \right)\). Note that ω is given by: \(\omega = \left \{ \begin {array}{ll} {\kappa _{M + 1}}, &if\ \xi = 1,\\ \frac {{{\kappa _2}}}{{{{\left ({1  \xi } \right)}^{M  1}}}}  \sum \limits _{m = 0}^{M  2} {\frac {{{\xi ^{1 + \rho }}{\kappa _{m + 2}}}}{{{{\left ({1  \xi } \right)}^{M  m  1}}}}},&otherwise. \end {array} \right.\)
Proof
Please refer to Appendix 1. □
Remarks 1
From (7), it is easily observed that the closeform expression of the secrecy outage probability, \({P^{I}_{c,sop}}\), is negatively correlated with the base station density λ _{ b } and the D2D transmitter density λ _{ d }. In contrast, it is positively correlated with the eavesdropper density λ _{ e }. This is due to the fact that the average received aggregate interference to confuse the most detrimental eavesdropper will be stronger with the increase of λ _{ b } or λ _{ d }. However, the average received SINR at the most detrimental eavesdropper will be higher as λ _{ e } increases. Furthermore, the detailed reason why the secrecy outage probability decreases as λ _{ b } increases is given by the following Corollary 1.
In addition, from (7) we can easily know that λ _{ b } only affects μ. Hence, we can obtain the following corollary.
Corollary 1
The secrecy outage probability of the cellular link is monotonically nonincreasing as λ _{ b } increases and it is independent of λ _{ b } when λ _{ b } is large enough.
This corollary implies that more base stations could improve the secrecy performance of the cellular link. This is because that the average received aggregate interference at each eavesdropper can be shown to scale with the base station density as (λ _{ b } p _{ a })^{ α/2}. Since eavesdroppers follows HPPP on a twodimensional plane, the received signal power at the most detrimental eavesdropper scales less than (λ _{ b } p _{ a })^{ α/2}. Hence, the secrecy outage probability is negatively correlated with the base station density. However, from (4), we can know that λ _{ b } p _{ a } will approach λ _{ u } when the base station density is large enough. Therefore, in this case, the secrecy outage probability will be independent of λ _{ b }.
Corollary 2
Proof
Following from Theorem 1 by letting N _{0}→0. □
From (8), we can see that there are close relationships between the secrecy outage probability of the cellular link and some main system parameters, such as the number of antennas M, the power allocation ratio ϕ. To evaluate the effect of ϕ and M on the secrecy outage performance, next we will derive the asymptotic expression of \({P^{I}_{c,sop}}\) when the number of antennas at each base station approaches infinity. We firstly give the following lemma when the number of antennas approaches infinity^{3}.
Lemma 1
\(\mathop {\lim }\limits _{M \to \infty } {\left \ {{\mathbf {g}}_{0e}^{T}{{\mathbf {W}}_0}} \right \^{2}} = M  1\), \(\mathop {\lim }\limits _{M \to \infty } {\left \ {{\mathbf {g}}_{ie}^{T}{{\mathbf {W}}_i}} \right \^{2}} = M  1\).
Proof
We can easily obtain Lemma 1 due to the fact that \({\left \ {{\mathbf {g}}_{0e}^{T}{{\mathbf {W}}_0}} \right \^{2}}\sim {\text {Gamma}}\left ({M  1,1} \right)\), \({\left \ {{\mathbf {g}}_{ie}^{T}{{\mathbf {W}}_i}} \right \^{2}} \sim \text {Gamma}\left ({M  1,1} \right)\). □
where \(I_{e\backslash \left \{ 0 \right \}}^{\infty } = \sum \limits _{{x_i} \in {\Phi ^{a}_b}\backslash \left \{ 0 \right \}} {\left ({{p_I}{{\left  {{\mathbf {g}}_{ie}^{T}{{\mathbf {w}}_i}} \right }^{2}}{{+ }}{p_A}} \right){{\left \ {{x_i}  {x_z}} \right \}^{ \alpha }}} + \sum \limits _{{y_i} \in {\Phi _d}} {{p_d}{{\left  {{h_i}} \right }^{2}}{{\left \ {{y_i}  {x_z}} \right \}^{ \alpha }}}\) denotes the cumulative interference from legitimate transmitters when the number of antennas at each base station approaches infinity. For notational conciseness, we define \(I_{_{e,c  e}}^{\infty } = \sum \limits _{{x_i} \in {\Phi ^{a}_b}\backslash \left \{ 0 \right \}} {\left ({{p_I}{{\left \ {{\mathbf {g}}_{ie}^{T}{{\mathbf {w}}_i}} \right \}^{2}}{{+ }}{p_A}} \right)} {\left \ {{x_i}  {x_z}} \right \^{ \alpha }}\), which similarly denotes the cumulative interference induced by both the informationbearing signal and the artificial noise from other base stations equipped with infinity antennas (except the typical base station). \(I_{_{e,d  e}}^{\infty } = \sum \limits _{{y_i} \in {\Phi _d}} {{p_d}{{\left  {{h_i}} \right }^{2}}{{\left \ {{y_i}  {x_z}} \right \}^{ \alpha }}}\). Thus, we can obtain \(I_{e\backslash \left \{ 0 \right \}}^{\infty } = I_{_{e,c  e}}^{\infty } + I_{_{e,d  e}}^{\infty } \).
Then, we have the following proposition.
Proposition 1
where we define \(\upsilon _{1} =2\pi {\lambda _e}{e^{ {{\hat \gamma }_e}\left ({M  1} \right)\xi }}\) and \(\mu _{1} = \left ({\pi {\lambda _b}{p_a}\Gamma \left ({1  \rho } \right)\Psi p_{I}^{\rho } + \frac {{\pi {\lambda _d}p_{d}^{\rho } }}{{\sin c\rho }}} \right){s^{\rho }}\phantom {\dot {i}}\) for notational conciseness. Note that \(\Psi = \Gamma \left ({1 + \rho,\left ({{\phi ^{ 1}}  1} \right)} \right){e^{\left ({{\phi ^{ 1}}  1} \right)}}\phantom {\dot {i}}\).
Proof
Please refer to Appendix 2. □
Since ξ=(ϕ ^{−1}−1)/(M−1), we can easily obtain that the asymptotic secrecy outage probability is independent of the number of antennas from Proposition 1.
By letting N _{0}→0, it is straightforward to obtain the following corollary for the interferencelimited case when the number of antennas at each base station approaches infinity.
Corollary 3
Proof
Following from Proposition 1 by letting N _{0}→0. □
When eavesdroppers have no multiuser decedability, even if each base station has no transmission power to generate the artificial noise, the intercell interference induced by the cellular link and the intracell interference induced by the D2D link could also confuse eavesdroppers. Hence, the base station is unnecessary to inject the artificial noise at some specified conditions. Based on the analysis above, we can obtain the following corollary employing the asymptotic expression of \({P^{I,asy,int}_{c,sop}}\) in the interferencelimited network when the number of antennas at each base station approaches infinity.
Corollary 4
where ε represents the minimum secrecy requirement for the cellular link.
Proof
Since \({P^{I,asy,int}_{c,sop}}\) is a monotonic increasing function with respect to the power allocation ratio ϕ. The secrecy performance of the cellular link would be satisfied as long as the secrecy outage probability is no more than ε, which is determined by the value of ϕ. By substituting ϕ=1 into the above constraint, we can obtain Eq. (12) given by Corollary 4. This provides very useful insight for practical system designs. □
3.1.2 Scenario II
where \(\frac {{{p_A}}}{{M  1}}{{\left \ {{\mathbf {g}}_{0e}^{T}{{\mathbf {W}}_0}} \right \}^{2}}{{\left \ {{x_z}} \right \}^{ \alpha }}\) means the received interference induced by the injected artificial noise from the typical base station. \(I_{A{\backslash \left \{ 0 \right \}}}={\sum \limits _{{x_i} \in {\Phi ^{a} _b}\backslash \left \{ 0 \right \}} {\frac {{{p_A}}}{{M  1}}{{\left \ {{\mathbf {g}}_{ie}^{T}{{\mathbf {W}}_i}} \right \}^{2}}} }{{\left \ {{x_i}  {x_z}} \right \}^{ \alpha }}\) denotes the cumulative interference from other base stations (except the typical base station) induced by the artificial noise. Then, we will give the expression of the secrecy outage probability of the cellular link in this case in Theorem 2.
Theorem 2
Note that \(\upsilon _{2} = 2\pi {\lambda _e} {\left ({1 + {{\hat \gamma }_e}\xi } \right)^{1  M}}\) and μ _{2}=λ _{ b } p _{ a } C _{ ρ,M } Θ ^{ ρ } r ^{2} are defined for notational conciseness, where \(\Theta ={\frac {{{{\hat \gamma }_e}\left ({1  \phi } \right)}}{{\left ({M  1} \right)\phi }}}\) and \({C_{\rho,M}} = \pi \frac {{\Gamma \left ({M  1 + \rho } \right)\Gamma \left ({1  \rho } \right)}}{{\Gamma \left ({M  1} \right)}}\).
Proof
Please refer to Appendix 3. □
Theorem 2 implies that the secrecy outage probability is negatively correlated with the base station density λ _{ b }. In contrast, it is positively correlated with the eavesdroppers density λ _{ e }. This stated remark agrees well with the remark from Theorem 1. Nevertheless, it is independent of the D2D transmitters density λ _{ d }. This is because that eavesdroppers have the multiuser decedebility to remove the interference induced by the D2D link and thus result in no impact on the eavesdropping link.
In addition, we can easily obtain that \(p^{II}_{c,sop}\) increases as ϕ increases, which denotes the power allocation ratio of the total transmission power allocated to the information transmission power. This is because that only the artificial noise will confuse eavesdroppers in this worst case. While a higher ϕ represents a lower transmission power allocated to generate the artificial noise to confuse eavesdroppers at the base station. Therefore, this will result in a much higher secrecy outage probability with a higher ϕ.
Corollary 5
To evaluate the effect of ϕ and M on the secrecy performance, similar to Proposition 1, next we will derive the asymptotic expression of the secrecy outage probability when eavesdroppers have the multiuser decedability and each base station has infinity antennas.
Proposition 2
where \({\upsilon _{2} = 2\pi {\lambda _e} {e^{ {{\hat \gamma }_e}\left ({M  1} \right)\xi }}}\) and μ _{3}=π λ _{ b } p _{ a } Γ(1−ρ)(ϕ ps)^{ ρ }(ϕ ^{−1}−1)^{ ρ } are defined for notational conciseness.
Proof
where p _{ A }∥x _{ z }∥^{−α } represents the received interference induced by the injected artificial noise from the typical base station equipped with infinity antennas. \(I^{\infty }_{A{\backslash \left \{ 0 \right \}}}={\sum \limits _{{x_i} \in {\Phi ^{a} _b}\backslash \left \{ 0 \right \}} {{{{p_A}}}{{\left \ {{x_i}  {x_z}} \right \}^{ \alpha }}} }\) represents the cumulative interference from other base stations with infinity antennas (except the typical base station) induced by the artificial noise. □
Then, substituting (19), (18) into (5) and changing to a polar coordinate system to evaluate the integral, we can obtain the result in (16).
We can also easily observe that the asymptotic expression of the secrecy outage probability of the cellular link is independent of the number of antennas, M, which agrees with the conclusion drawn from Proposition 1.
Then, it is straightforward to obtain the following corollary by letting N _{0}→0.
Corollary 6
Note that we define \(\eta ={{\pi {\lambda _b}{p_a}\Gamma \left ({1  \rho } \right)\left ({\phi ^{1}1} \right)^{\rho }{{{{\hat \gamma }_e}}^{\rho } }}}\) for notational conciseness.
3.2 Connection outage probability of D2D links
where p _{ d } is the transmission power of all D2D transmitters. \({{I_d} \,=\, \sum \limits _{{x_i} \in {\Phi ^{a} _b}} { p_{I}{{\left \ {{x_i}} \right \}^{ \alpha }}{{\left  {{\mathbf {g}}_{id}^{T}{{\mathbf {w}}_i}} \right }^{2}}}\,+\, \sum \limits _{{x_i} \in {\Phi ^{a} _b}} {\frac {{p_A}}{{M  1}}{{\left \ {{x_i}} \right \}^{ \alpha }}}} {{{\left \ {{\mathbf {g}}_{id}^{T}{{\mathbf {W}}_i}} \right \}^{2}} + \sum \limits _{{y_i} \in {\Phi _d}\backslash \left \{ {{y_0}} \right \}} {{p_d}{h_{id}}{{\left \ {{y_i}} \right \}^{ \alpha }}}}\) represents the totally cumulative interference from the base station that is located at x _{ i } and other D2D transmitters (except the typical D2D transmitter located at y _{0}). \({I_{c  d}} = \sum \limits _{{x_i} \in \Phi _{b}^{a}} {\left ({{p_I}{{\left  {{\mathbf {g}}_{i}^{T}{{\mathbf {w}}_i}} \right }^{2}} + \frac {{{p_A}}}{{M  1}}{{\left \ {{\mathbf {g}}_{i}^{T}{{\mathbf {W}}_i}} \right \}^{2}}} \right){{\left \ {{x_i}} \right \}^{ \alpha }}} \) represents the interference induced by the informationbearing signal and the artificial noise from all the base station. \({I_{d  d}} = \sum \limits _{{y_i} \in {\Phi _d}\backslash \left \{ {{y_0}} \right \}} {{p_d}{h_{id}}{{\left \ {{y_i}} \right \}^{ \alpha }}}\) represents the interference from other D2D links sharing the same resource except the typical D2D transmitter. h _{ id } represents the smallscale fading channel from the D2D transmitter located at y _{ i }, especially, h _{0d } means the smallscale fading channel from the typical D2D transmitter. Similarly, g _{ id } represents the smallscale fading channel from the base station located at x _{ i } to the typical D2D receiver. It is assumed that h _{ id } follows the exponential distribution with unit mean, i.e., h _{ id }∼ exp(1). N _{0} represents the covariance of the additive Gaussian noise at the typical D2D receiver. Hence, we can have I _{ d }=I _{ c−d }+I _{ d−d }.
where \({\hat \gamma _d}= 2^{R_d}1\) represents the SINR target threshold to satisfy the communication requirement of the typical D2D link and \(\zeta = \frac {{{{\hat {\gamma }_d}}{l^{\alpha } }}}{{{p_d}}}\). \({{{\mathcal L}}_{{I_d}}}\left (\zeta \right)\) denotes the Laplace transform of I _{ d }, i.e., \( {{{\mathcal {L}}}_{{I_d}}}\left (\zeta \right) = \mathbb {E}\left ({  \zeta {I_d}} \right)\). According to the property of the Laplace transform, we can easily have \( {{{\mathcal L}}_{{I_d}}}\left (\zeta \right) = {{{\mathcal L}}_{I_{c  d}}}\left (\zeta \right) \bullet {{{\mathcal L}}_{{I_{dd}}}}\left (\zeta \right)\) because of I _{ d }=I _{ c−d }+I _{ d−d }.
Theorem 3
Note that \({k \,=\, \left \{ \begin {array}{l} 2p_{I}^{\rho } {\kappa _{M + 1}}, \qquad \qquad \qquad \qquad \quad \;\,\, if\ \xi = 1, \\ 2p_{I}^{\rho } \left ( {\frac {{{\kappa _2}}}{{{{\left ( {1  \xi } \right)}^{M  1}}}} \,\, \sum \limits _{m = 0}^{M  2} {\frac {{{\xi ^{1 + \rho }}{\kappa _{m + 2}}}}{{{{\left ({1  \xi } \right)}^{M  m  1}}}} } } \right), otherwise. \end {array} \right.}\)
Proof
Please refer to Appendix 4. □
Remarks 2
From the derived result in (23), it is obvious that the connection outage probability, p _{ d,cop }, has close relationships with various system parameters, such as the densities λ _{ b }, λ _{ e }, the total transmission power p of each base station, the D2D transmission power p _{ d } and so on. Especially, for the given λ _{ b }, λ _{ e } and ϕ, p _{ d,cop } is positively correlated with the transmission power ratio p / p _{ d }. This is because that a larger transmission power of each base station will introduce stronger interference to the typical D2D link from the cellular downlink, resulting in a larger connection outage probability of the typical D2D link.
Remarks 3
The expression of the connection outage probability in (23) is derived just under the assumption that the distance l between the typical D2D transmitter and its corresponding D2D receiver is constant. The derived result can be easily expanded to the scenario where l is a random variable. The expression of the connection outage probability of the typical D2D link in the expanded scenario can be obtained by calculating the integral formula \(\int _{0}^{\infty } \mathbb {P} \left ({\left. {\text {SINR}_{d} < {\beta _d}} \right l} \right){f_l}\left (l \right)dl\), where f _{ l }(l) denotes the PDF of the distance l.
By letting N _{0}=0, we will get the expression of the connection outage probability shown in the following corollary for the interferencelimited network.
Corollary 7
where k is given in (23).
Proof
Corollary 7 can be straightforwardly obtained from Theorem 3 with N _{0}→0. □
Similarly, next we will provide the asymptotic expression of the connection outage probability of the typical D2D link when each base station has infinity antennas.
Proposition 3
Proof
where \(I_{d}^{\infty } = \sum \limits _{{x_i} \in {\Phi ^{a}_b}} {\left ({{p_I}{{\left  {{\mathbf {g}}_{id}^{T}{{\mathbf {w}}_i}} \right }^{2}} + {p_A}} \right){{\left \ {{x_i}} \right \}^{ \alpha }}} + \sum \limits _{{y_i} \in {\Phi _d}\backslash \left \{ {{y_0}} \right \}} {{p_d}{h_{id}}{{\left \ {{y_i}} \right \}^{ \alpha }}}\) represents the totally cumulative interference when each base station has infinity number of antennas. \(I_{d,c  d}^{\infty } = \sum \limits _{{x_i} \in {\Phi ^{a}_b}} {\left ({{p_I}{{\left  {{\mathbf {g}}_{id}^{T}{{\mathbf {w}}_i}} \right }^{2}} + {p_A}} \right){{\left \ {{x_i}} \right \}^{ \alpha }}}\) represents the cumulative interference from all the base stations, \({I_3} = \sum \limits _{{y_i} \in {\Phi _d}\backslash \left \{ {{y_0}} \right \}} {{p_d}{h_{id}}{{\left \ {{y_i}} \right \}^{ \alpha }}}\) represents the cumulative interference from other D2D transmitters. Hence, it is intuitive that \(I_{d}^{\infty } = I_{d,c  d}^{\infty } + {I_3}\). □
where Ψ is given by (10).
Substituting (27), (41) into (22) could yield the result in (25). Then, we can easily get the following corollary from Proposition 3.
Corollary 8
Proof
Following from Proposition 3 by letting N _{0}→0. □
4 Numerical results and analysis
In this suction, more detailed simulation and numerical results are provided to evaluate the theoretical analysis. The path loss exponent is α=3 and μ=3.5 for the nearest base station association. The total transmission power of each base station is 60 dBm and the transmission power of all the D2D transmitters is 20 dBm. The density of the cellular user is 0.0005/m ^{2}, i.e., λ _{ c }=0.0005/m ^{2}. The number of antennas equipped at each base station is set to be M=10. For simplicity, it is assumed that the distance between D2D pairs is l=1m.
5 Conclusion
In this paper, secure communication for the cellular downlink is investigated in this hybrid network. A case was considered, in which each base station has no CSI from D2D users because they are generally deployed in the cell edge. To guarantee secure communication of the cellular link, each base station employed the artificial noise assisted transmission strategy. Firstly, we considered two different scenarios depending on whether eavesdroppers have the multiuser decedability or not and derived the closeform expression of the secrecy outage probability of the cellular link. To characterize the reliable communication of the D2D link, its closeform expression of connection outage probability was derived and some comprehensive analysis were provided to guide the system design. Finally, simulation results are provided to validate the effective of the theoretical results. Furthermore, more complex D2D scenes need to be studied.
6 Appendix 1: Proof of Theorem 1
Let us define \({\gamma _{0,{e}}} = {p_I}\left ({{{\left  {{\mathbf {g}}_{0e}^{T}{{\mathbf {w}}_0}} \right }^{2}}  \xi {{\hat \gamma }_e}{{\left \ {{\mathbf {g}}_{0e}^{T}{{\mathbf {W}}_0}} \right \}^{2}}} \right)\).
where ω is given by (7).
where \(\frac {1}{{\sin c\rho }}=\frac {{\pi \rho }}{{\sin \pi \rho }} =\Gamma \left ({1 + \rho } \right)\Gamma \left ({1  \rho } \right)\).
Substituting (34) into (5) and changing to a polar coordinate system to evaluate the integral yields the result in (7).
7 Appendix 2: Proof of Proposition 1
where Ψ is given by (10).
Substituting (36) into (5) and changing to a polar coordinate system to evaluate the integral yields the results in (10).
8 Appendix 3: Proof of Theorem 2
where \({{{\mathcal L}}_{{I_{A\backslash \left \{ 0 \right \}}}}}\left (s \right)\) denotes the Laplace transform of I _{ A∖{0}}, i.e., \( {{{\mathcal L}}_{{I_{A\backslash \left \{ 0 \right \}}}}}\left (s \right) = \mathbb {E}\left ({  s{{I_{A\backslash \left \{ 0 \right \}}}}} \right)\).
Then, substituting (38), (37) into (5) and changing to a polar coordinate system to evaluate the integral, we can get the result in (14).
9 Appendix 4: Proof of Theorem 3
where k is given in (23).
Then, combing (43), (44) and plugging into (22), we can obtain the result in (23).
The performance metrics to characterize the secrecy performance from different perspectives in existing works can be classified into two types: the secrecy outage probability and the achievable secrecy rate. The ergodic achievable secrecy rate is not suitable for the system having strictly realtime requirements. However, in the 5G mobile communication, it has higher realtime requirements. Hence, we focus on the secrecy outage probability to depict the secrecy performance of the cellular link in this paper.
In this paper, we adopt the TDMA scheme to derive the results and discuss the performance of this hybrid network, but the derived results can be easily expanded to other systems, such as the frequencydivision multiple access (FDMA) scheme and so on.
For the nullspace based beamforming, the matrix inversion in massive MIMO systems will incur very high computation cost. However, alternatively, we may use the random artificial noise scheme which has been adopted in [30]. From the derived results and numerical results [30], we can come to a conclusion that main system parameters have the same effect on the secrecy performance for different design schemes. The conclusions drawn from the derived result in this paper could also guide the system design when the artificial noise is in random form.
Notes
Declarations
Acknowledgements
This work is supported in part by China’s HighTech R&D Program (863 Program) SS2015AA011306; the open research fund of National Mobile Communications Research Laboratory, Southeast University (No.2013D09) and National Natural Science Foundation of China under Grants No.61379006, 61521003, and 61401510.
Authors’ contributions
YC put forward the idea and wrote the manuscript. XJ and KH took part in the discussion and they also guided, reviewed, and checked the writing. JY, XH, and YX carried out experiments and analyzed experimental results. All authors read and approved the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
 M Agiwal, A Roy, N Saxena, Next Generation 5G Wireless Networks: A Comprehensive Survey. IEEE Commun. Surv. Tutor. 18(3), 1617–1655 (2016).View ArticleGoogle Scholar
 A Asadi, Q Wang, V Mancuso, A survey on devicetodevice communication in cellular networks. IEEE Commun. Surv. Tutor. 16(4), 1801–1819 (2014).View ArticleGoogle Scholar
 G Ding, J Wang, Q Wu, Q Yao, Y Song, F Tsiftsis, Cellularbasestationassisted devicetodevice communications in TV white space. IEEE J. Sel. Areas Commun. 34(3), 107–121 (2016).View ArticleGoogle Scholar
 Technical specification group services and system aspects; feasibility study for proximity services (ProSe). Cedex, France, 3GPP TR 22.803, 2012, Rel12.Google Scholar
 LS on agreements from TSG RAN on work on public safety related use cases in Release 12. Cedex, France, 3GPP TD SP130478, Sep. 2013.Google Scholar
 J Yue, C Ma, H Yu, W Zhou, Secrecybased access control for devicetodevice communication underlaying cellular networks. IEEE Commun. Lett. 17(11), 2068–2071 (2013).View ArticleGoogle Scholar
 R Zhang, X Cheng, L Yang, Cooperation via spectrum sharing for physical layer security in devicetodevice communications underlaying cellular networks. IEEE Trans. Wirel. Commun. 15(8), 5651–5663 (2016).View ArticleGoogle Scholar
 L Sun, Q Du, P Ren, Y Wang, Two birds with one stone: Towards secure and interferencefree D2D transmissions via constellation rotation. IEEE Trans. Veh. Technol. 65(10), 8767–8774 (2016).View ArticleGoogle Scholar
 X Kang, X Ji, K Huang, X Li, Securityoriented distributed access selection for D2D underlaying cellular networks. IET Electron. Lett. 53(1), 32–34 (2017).View ArticleGoogle Scholar
 Y Chen, X Ji, K Huang, X Kang, Secrecyoutageprobabilitybased Access Strategy for Devicetodevice Communications Underlaying Cellular Networks. J. Commun. 37(8), 86–94 (2016).Google Scholar
 Y Chen, X Ji, K Huang, B Li, X Kang, Opportunistic access control for enhancing security in D2Denabled cellular networks. Sci China Inf Sci (2016). doi:10.1007/s114320179160y.
 Y Liu, L Wang, S Zaidi, M Elkashlan, T Duong, Secure D2D Communication in LargeScale Cognitive Cellular Networks: A Wireless Power Transfer Model. IEEE Trans. Commun. 64(1), 329–342 (2016).View ArticleGoogle Scholar
 C Ma, J Liu, X Tian, H Yu, Y Cui, X Wang, Interference exploitation in D2Denabled cellular networks: A secrecy perspective. IEEE Trans. Commun. 63(1), 229–242 (2015).Google Scholar
 Z Chu, K Cumanan, M Xu, Z Ding, Robust secrecy rate optimizations for multiuser multipleinputsingleoutput channel with devicetodevice communications. IET Commun.9(3), 396–403 (2015).View ArticleGoogle Scholar
 Z Chu, X Nguyen, T Le, M Karamanoglu, et al., Game theory based secure wireless powered D2D communications with cooperative jamming, (2017 Wireless Days, Porto, 2017).Google Scholar
 S Goel, R Negi, Guaranteeing secrecy using artificail noise. IEEE Trans. Wirel. Commun. 7(6), 2180–2189 (2008).View ArticleGoogle Scholar
 Z Chu, Z Zhu, M Johnston, et al., Simultaneous Wireless Information Power Transfer for MISO Secrecy Channel. IEEE Trans. Veh. Technol. 65(9), 6913–6925 (2016).View ArticleGoogle Scholar
 X Kang, X Ji, K Huang, Secure D2D Underlaying Cellular Communication Based on Artificial Noise Assisted. J. Commun. 36(10), 149–156 (2015).Google Scholar
 X Kang, X Ji, K Huang, Z Zhong, Secure D2D communication Underlaying Cellular Networks: Artificial Noise Assisted. in Proceedings of the IEEE International Conference on Vehicular Technology (VTC). (Montreal, 2016).Google Scholar
 A Thangaraj, S Dihidar, AR Calderbank, SW McLaughlin, JM Merolla, Applications of LDPC codes to the wiretap channel. IEEE Trans. Inf. Theory. 53(8), 2933–2945 (2007).MathSciNetView ArticleMATHGoogle Scholar
 X Xu, B He, W Yang, X Zhou, Y Cai, Secure Transmission Design for Cognitive Radio Networks With Poisson Distributed Eavesdroppers. IEEE Trans. Inf. Forensic Secur. 11(2), 373–387 (2016).View ArticleGoogle Scholar
 C Li, J Zhang, KB Letaief, Throughput and energy efficiency analysis of small cell networks with multiantenna base stations. IEEE Trans. Wirel. Commun. 13(5), 2505–2517 (2014).View ArticleGoogle Scholar
 C Liu, L Wang, Optimal cell load and throughput in green small cell networks with generalized cell association. IEEE J. Sel. Areas Commun. 34(5), 1058–1072 (2016).View ArticleGoogle Scholar
 D Stoyan, W Kendall, J Mecke, Stochastic Geometry and Its Applications, 2nd ed (Wiley, Hoboken, 1996).MATHGoogle Scholar
 X Zhang, X Zhou, MR McKay, Enhancing secrecy with multiantenna transmission in wireless ad hoc networks. IEEE Trans. Inf. Forensic Secur. 8(11), 1802–1814 (2013).View ArticleGoogle Scholar
 H Wang, T Zheng, J Yuan, D Towsley, M Lee, Physical Layer Security in Heterogeneous Cellular Networks. IEEE Trans. Commun. 64(3), 1204–1219 (2016).View ArticleGoogle Scholar
 W Wang, KC Teh, KH Li, Artificial Noise Aided Physical Layer Security in MultiAntenna SmallCell Networks. IEEE Trans. Inf. Forensic Secur. 12(6), 1470–1483 (2017).View ArticleGoogle Scholar
 I Gradshteyn, I Ryzhik, A Jeffrey, D Zwillinger, S Technica, Table of Integrals, Series, and Products (Academic, New York, 2007).Google Scholar
 M Haenggi, JG Andrews, F Baccelli, O Dousse, M Franceschetti, Stochastic geometry and random graphs for the analysis and design of wireless networks. IEEE J. Sel. Areas Commun. 27(7), 1029–1046 (2009).View ArticleGoogle Scholar
 J Zhu, R Schober, V Bhargava, Secure transmission in multicell massive MIMO systems. IEEE Trans. Wirel. Commun. 13(9), 4766–4781 (2014).View ArticleGoogle Scholar