Joint Cooperative Relaying and Jamming for Maximum Secrecy Capacity in Wireless Networks

Transcription

1 Joint Cooperative Relaying an Jamming for Maximum Secrecy Capacity in Wireless Networks Li Wang, Chunyan Cao, Mei Song an Yu Cheng Beijing Key Laboratory of Work Safety Intelligent Monitoring School of Electronic Engineering, Beijing University of Posts an Telecommunications, Beijing, P.R. China Department of Electrical an Computer Engineering Illinois Institute of Technology 330 S. Dearborn Street, Siegel Hall 320 Chicago, IL Abstract This paper proposes a joint cooperative relaying an jamming scheme base on istribute beamforming to enhance the secrecy rate of a wireless channel in the presence of an eavesropper. Specifically, we consier the scenario that a source noe transmits messages to a estination with the help of multiple cooperative relays, where a relay might be compromise to become an eavesropper as an insie attacker. Our joint relaying an jamming approach is to assign some relay noes to act as jammers to interfere the eavesropper, while the remaining ones continue relaying information to the estination. We propose a protocol to implement the joint cooperative relaying an jamming. Our protocol further consiers the particular challenge brought by the insie attacker: it may know the jamming signal an can use it to remove the interference from the jammers. The issue of phase an frequency synchronization in the istribute beamforming is also taken into account. A mixe integer programming problem is formulate to maximize the secrecy rate with the propose joint relaying an jamming, by which the optimal role assignment to a relay an the associate optimal power assignment can be solve. Numerical results are presente to emonstrate the efficiency of the propose joint relaying an jamming in improving the secrecy rate, with comparison to existing approaches. Inex Terms Cooperative relaying an jamming, istribute beamforming, synchronization, insie attacker, secrecy rate. I. INTRODUCTION Recently, physical layer security has been regare as a promising approach to aress security issues in wireless networks without upper-layer encryption, which exploits the physical characteristics of the wireless channels to ensure secure communications. Secrecy rate is efine as the rate at which information can be transmitte secretly from the source to its intene estination in the presence of eavesroppers, an the maximum achievable secrecy rate is name the secrecy capacity. In [], Aaron Wyner introuce the wiretap channel an establishe the funamental results of creating perfect secure communications without cryptographic techniques involve. Wyner s approach was extene to Gaussian wiretap channels an broacast channels in later works [2]-[3]. Cooperative transmission has attracte a lot of attention to enhance the capacity of a wireless channel as a virtual MIMO system [4]-[5]. The current efforts to improve the secrecy rate in the context of cooperative communications can be roughly classifie into three categories, cooperative beamforming, cooperative jamming, an hybri relaying an jamming. For cooperative beamforming, all relays resort to the istribute beamforming technique to concentrate the signal towars the intene estination, trying to mitigate the information leakage to the eavesropper as much as possible [6]- [7], [8]. The philosophy of cooperative jamming is that some relay noes concentrate certain jamming signal to interfere the eavesropper (normally by the istribute beamforming technique too) while not impact the estination [9]-[0], []. A relay equippe with multiple antennas for jamming was stuie in [0]. However, the cost of a multiple antenna system is normally high. The work in [2] presente a systematic stuy of cooperative beamforming an cooperative jamming. The cooperative jamming situation consiere in [2] only applies to the one-hop source to estination transmission though. Hybri relaying an jamming schemes to secure the networks were propose in [3]-[4], where some noes aopt istribute beamforming to relay the message an others jam the eavesroppers. It is worth noting that most of existing works, in all the three categories, consiere a given set of relay(s), jammer(s) an the eavesropper(s), which may not be the case in practice: a normal noe coul be compromise to turn into an attacker an the role of a relay noe coul be ynamically assigne. This paper consiers a practical scenario where a transmission with multiple cooperative relays starts at the normal state an the relays can apply istribute beamforming technique for high capacity. A relay noe might be compromise to become an eavesropper, which is an insie attacker. We assign some relay noes to act as jammers to interfere the eavesropper, while the rest continue relaying information to the estination. The unique issue in our scenario is the optimal role assignment to each relay, as a ata relay or a jammer, for maximum secrecy rate. We formulate a mixe integer programming problem. Furthermore, we consier the particular challenge brought by the insie attacker: all the jammers nee to transmit the same jamming symbols an use the istribute beamforming technique for effective jamming, however, the insie attacker may also know the jamming symbols (if preloae) an can use it to remove the interferences from the jammers. We thus esign a protocol to aress such an issue. In aition, our protocol incorporates operations for /4/$ IEEE 4448

2 Fig.. System moel for secure transmission. synchronization in istribute beamforming, which is ignore in most of the existing stuies on secrecy rate for cooperative communications. To make our scheme more efficient an practical, we introuce two hop istribute beamforming for joint cooperative relaying an jamming. The paper is organize as follows. Section II presents the system moel. Section III presents the protocol esign for joint cooperative relaying an jamming. Section IV formulates a mixe integer programming problem to maximize the secrecy rate. Numerical results are provie in Section V to emonstrate the performance. Section VI gives conclusion remarks. II. SYSTEM MODEL In this section, we will first present the system moel an then give a brief escription of the secure transmission between the source an the estination. As shown in Fig., we consier the scenario where a source nees to communicate with a estination uner the help of some cooperative relays in the DF manner. Each noe is equippe with a single omni-irectional antenna an operates in a half-uplex moe. We assume that channels for ifferent pairs of noes are inepenent an ientically istribute (i.i..) with flat Rayleigh faing, an the noise is the aitive white Gaussian noise (AWGN) with a mean of zero an a variance of σ 2. In the system, a relay noe might be compromise to become an eavesropper, as an insie attacker, which can be etecte by its malicious behavior. Some relay noes will be assigne as jammers to interfere the eavesropper, while the remaining ones continue relaying information to the estination. Here we assume the eavesropper can be locate by some existing techniques [5], an all caniate noes can successfully ecoe the message transmitte by the source as assume in [2]. The secure transmission process between the source an the estination can be ivie into two transmitting phases. In the first phase, the source broacasts the message, S, which is also listene by the eavesropper. To ecrease the interception, the jammers sen jamming signals, Z, at the same time. Distribute beamforming is employe at the jammers so that the jamming signals confuse the eavesropper only, without impacting the estination. In the secon phase, the relays utilize istribute beamforming to forwar the ata to the estination, while the jammers continue sening jamming signals to interfere the eavesropper. When useful message sent by the relays can be completely nulle out at the eavesropper using null space cooperative beamforming at the relays [8], the jamming is not neee an the transmit power of the jammer set can be set zero. While if cooperative beamforming at the relays oes not work well, the jammers interfere the eavesropper an ecrease its interception. Optimal power allocation to jammers an relays will be stuie in Section IV. We enote the channel gain form the source to the estination by h s, the channel gain form the source to the eavesropper by h se, the row channel gain vector ( N) from the source to all the caniate noes by h T sn, the column channel gain vector (N ) from all the caniate noes to the eavesropper by h Ne, the column channel gain vector (N ) from all the caniate noes to the estination by h N, the channel gain matrix (N N) between all the caniate noes by H N. N is the number of caniate noes. ( ) T is the transpose of a matrix or a vector. [x] + =max(0,x). III. JOINT COOPERATIVE RELAYING AND JAMMING In this section, we esign a protocol to implement the propose cooperative relaying an jamming scheme. The technique unerpinning of the joint relaying an jamming is the istribute beamforming. An important issue with istribute beamforming is the synchronization of the carrier signals in phase an frequency. The stuy in [6] presente an effective synchronization technique for istribute beamforming an inicate that the beamforming gain still remains goo even with imperfect synchronization. We evelop our protocol base on that use in [6]. Furthermore, the severe impact of insie attacks from the eavesropper can not be ignore. Cooperative jamming relies on istribute beamforming where the jammers nee to transmit the same jamming symbols for effective jamming, however, the eavesropper may also know the preefine jamming symbols an can use it to remove the interference. Our protocol utilizes on-line selection of the jamming signals rather than pre-installation to be against the insie attacker. Although we involve some cryptographic techniques here to eliver the on-line selecte jamming signal from the source to the jamming noes, the cryptographic technique just incurs some overhea to our cooperative relaying an jamming, an the secrecy rate still epens on the physical layer security technique. The propose protocol works as follows. First, the source generates a sinusoi signal c 0 (t) with a frequency f c base on its local oscillator an broacasts it. Here we assume that the source has one pair of public/private keys. The sinusoi signal c 0 (t) can be moulate with the public key of the source. The jammers use PLL (Phase Locking Loop) to lock on the frequency f c an the receive signal at the κth jammer noe is c κ,0 (t) with a phase shift θ κ. The jammers apply phase error correction techniques to eal with phase shift θ κ [6]. Then the κth jammer noe obtains the calibrate signal c κ (t) use for channel estimation an beamforming. All the jammers also emoulate the public key 4449

3 of the source to be use later for jamming signal transmission. An easier (but not that general) implementation is that the public key of the source is pre-installe into each relay noe. Secon, all the caniate noes an the estination then take turns to broacast a carrier signal so that each noe can estimate the channel gain between itself an the broacasting one. These channel gains can later be use for istribute beamforming, either for cooperative relaying to the estination or for jamming to the eavesropper. Thir, if each jammer obtains the public key of the source in the first step, it then uses this key to encrypt its session key to the source (here we use the local phase error as a session key, which is relate to the propagation elay of the wireless channel between the source an this jammer). The source can later use the session key to sen upate jamming signals use for istribute beamforming to the jammers, which is similar to the metho in [7] by exploiting the wireless channel for builing cryptographic services. Last, at a certain moment, if the source announces the eavesropper, it will sen the upate jamming signals to each jammer, encrypte by their session keys, respectively. The jammers then start to jam the eavesropper with istribute beamforming. Since the on-line selecte jamming signal is encrypte by the session key, the eavesropper (the insie attacker) cannot obtain it. Please be note that the propose protocol can be implemente with a time-ivie system, by extening the frame structure presente in [6]. As the focus of this paper is to emonstrate the effectiveness of the joint cooperative relaying an jamming an the associate optimal relay role assignment, we will fully evelop the implementation etails an conuct relate overhea stuy in our future work. IV. SECRECY CAPACITY MAXIMIZATION In this section, we will first formulate the secrecy rate of the whole transmission, an present a mixe integer programming problem to maximize the secrecy rate, which relates to the optimal assignment of the roles to caniate noes. Distribute beamforming is utilize in the moel, where each relay noe an jammer noe apply a weight to the signal transmitte. With proper selection of the weights, the receive signal at a certain noe can be nulle out to achieve jamming through istribute beamforming. A. Secrecy Rate Formulation As inicate before, to improve the secrecy rate of the whole transmission, a two-phase transmission between the source an the estination is establishe. Each caniate noe can be regare as a relay or a jammer. We present an optimal relay an jammer selection to maximize the secrecy rate. Denote the set of all the caniate noes N = {, 2,,..., N}. For each caniate noe, we set two parameters to inicate its role as a relay or jammer. The specifical escription can be expresse as { relay if x i = an u i =0, n i = jammer if x i = 0 an u i =, where n i is the ith caniate noe (i =,..., N), the value of x i an u i can be 0 or, an x i + u i =. Accoring to the efinition, for convenience of escription an explicit physical meaning, we enote the roles of all N caniate noes by two vectors x =[x,x 2,..., x N ] T an u = [u,u 2,..., u N ] T. Therefore, we can obtain the role of the ith caniate noe by the values of x i an u i, which are relative to the secrecy rate maximization problem. ) The first transmitting phase: The source broacasts the message, S, an certainly the eavesropper can hear the message. To isturb the eavesropping, the jammers at the same time sen the jamming signals, Z, by istribute beamforming without affecting the estination. For the caniate noes acting as relays, the values corresponing to them in vector x must be, an the values corresponing to the caniate noes acting as jammers in vector u must be. In this phase, the receive signals at the caniate noes, the estination an the eavesropper can be written as y N = P s h T sn S + n N, () y () = P s h s S + n (), (2) y e () = P s h se S + P u T W h Ne Z + n () e, (3) respectively, where P s is the transmit power of the source, P is the total transmit power of the caniate noes in the first phase, W = iag([w (), w() 2,..., w() N ]) is a N N power allocation weight matrix of the caniate noes in the first phase, w T w = where w =[w (), w() 2,..., w() N ]T, n N is the AWGN at the caniate noes, n () is the AWGN at the estination an n () e is the AWGN at the eavesropper. The receive SNRs at the estination an the eavesropper in the first phase can be escribe as γ () = P s h s 2 σ 2, (4) γ e () P s h se 2 = P u T W h Ne 2 + σ 2, (5) respectively. 2) The secon transmitting phase: The relays utilize istribute beamforming to forwar the weighte message to the estination while the jammers continue sening jamming signals to interfere the eavesropper without impacting the estination. In this phase, the receive signals at the estination an the eavesropper can be written as y (2) = P 2 x T W 2 h N S + n (2), (6) y e (2) = P 2 x T W 2 h Ne S + P 2 u T W 2 h Ne Z + n (2) e, (7) respectively, where P 2 is the total transmit power of the caniate noes in the secon phase, W 2 = iag([w (2), w(2) 2,..., w(2) N ]) is a N N power allocation weight matrix of the caniate noes in the secon phase, w2 T w 2 = where w 2 =[w (2), w(2) 2,..., w(2) N ]T, n (2) is the AWGN at the estination an n (2) e is the AWGN at the eavesropper. 4450

4 The receive SNRs at the estination an the eavesropper in the secon phase can be escribe as γ (2) = P 2 x T W 2 h N 2 σ 2, (8) γ e (2) = P 2 x T W 2 h Ne 2 P 2 u T W 2 h Ne 2 + σ 2, (9) respectively. In the secon phase, when the esire message sent by the relays can be completely nulle out at the eavesropper by using cooperative beamforming at the relays, i.e., x T W 2 h Ne =0, the eavesropper can wiretap nothing about the esire message, then the jamming is not necessary. We combine the signals from the first an secon phases aopting maximal ratio combining (MRC) metho [8], an the receive SNRs at the estination an the eavesropper in the whole transmission can be escribe as γ = γ () + γ (2) = P s h s 2 σ 2 + P 2 x T W 2 h N 2 σ 2, (0) γ e = γ e () + γ e (2) P s h se 2 = P u T W h Ne 2 + σ 2 + P 2 x T W 2 h Ne 2 P 2 u T W 2 h Ne 2 + σ 2, () respectively. When x T W 2 h Ne =0, the secon part of the above equation is zero, i.e., γ e (2) =0. The corresponing maximum information rates at the estination an the eavesropper are R = 2 log 2( + γ )=log 2 ( + γ () + γ (2) ), (2) R e = 2 log 2( + γ e )=log 2 ( + γ () e + γ (2) e ), (3) respectively, where the factor /2 is ue to the fact that the two phases share the transmission time. Therefore, the secrecy rate of the whole transmission can be written as SC = {R R e } + = 2 {log 2 ( + γ ) log 2 ( + γ e )} +. (4) B. Maximization Problem From the above iscussion, we can see that changing the number of relays an jammers will impact the secrecy rate, i.e., ifferent values of the vectors x an u influence the security of the transmission. Thus, a mixe integer programming problem for secrecy rate maximization with the optimal values of the vectors x an u is presente. In the first transmitting phase, istribute beamforming is aopte at the jammers to avoi the interference at the relays an the estination while confusing the eavesropper. Thus, there are two null space constraints in this phase, i.e., u T W H N X = 0 N an u T W h N = 0, where X = iag([x, x 2,..., x N ]). The necessary conition for the null space constraints in this phase is that the number of jammers is larger than that of relays an estination, i.e., N J >N R +, where N J an N R are the numbers of jammers an relays, respectively. Otherwise, if N J N R +, we cannot avoi interference cause by the jammers at the relays an the estination completely. If cooperative beamforming at the relays oes not work well in the secon transmitting phase, the jammers interfere the eavesropper without affecting the estination, that is, u T W 2 h N =0an N J >. When null space cooperative beamforming employe at the relays can completely null out the esire message at the eavesropper, then x T W 2 h Ne =0 an N R >. Our objective is to select optimal relays an jammers with power allocation to maximize the secrecy rate subject to a total transmit power constraint. Depening on whether x T W 2 h Ne is zero or not, we can formulate the secrecy rate maximization problem as following. ) x T W 2 h Ne 0 (without nulling out): The esire message sent by the relays cannot be completely nulle out at the eavesropper in the secon phase. Uner the null space constraints in the first phase, the optimization problem for maximum secrecy rate can be formulate as follows. arg max SC =arg max {R R e } =arg max 2 {log 2 ( + γ ) log 2 ( + γ e )}, (5) subject to P s + P + P 2 = P 0, (5a) u T W h N =0, (5b) u T W 2 h N =0, (5c) u T W H N X = 0 N, (5) where P 0 is the total transmit power constraint. It is obviously a mixe integer programming problem. Specifically, when SNR, the objective function can be expresse as arg max 2 {log 2 ( + γ ) log 2 ( + γ e )} arg max P s h s 2 σ + P2 xt W 2h N 2 2 σ 2 P s h se 2 P u T W h Ne + P2 xt W 2h Ne 2 2 +σ 2 P 2 u T W 2h Ne 2 +σ 2. (6) 2) x T W 2 h Ne =0(nulling out): The esire message sent by the relays can be completely nulle out at the eavesropper in the secon phase. In this case, the secrecy rate maximization problem can be formulate as follows. arg max SC =arg max 2 {log 2 ( + γ ) log 2 ( + γ e )}, (7) subject to P s + P + P 2 = P 0, (7a) u T W h N =0, (7b) x T W 2 h Ne =0, (7c) u T W H N X = 0 N. (7) 445

5 propose scheme 0.2 DF scheme CJ scheme Source-eavesropper Distance (m) Fig. 2. Scenario use for numerical experiments. Fig. 3. Secrecy rate versus source-eavesropper istance. Similarly, when SNR, the objective function will be arg max 2 {log 2 ( + γ ) log 2 ( + γ e )} arg max =arg max P s h s 2 σ + P2 xt W 2h N 2 2 σ 2 P s h se 2 P u T W h Ne 2 +σ 2 (P s h s 2 + P 2 x T W 2 h N 2 ) P u T W h Ne 2 + σ 2 P s h se 2. (8) We can see that the secrecy rate maximization problem is a mixe integer programming problem, which is NP-har in general. In this paper, our focus is to emonstrate the efficiency of the propose joint relaying an jamming an the optimal relay assignment, an we use exhaustive search to solve the problem which is an easy approach to obtain the solution when the network scale is small, while it becomes ifficult for a large network scale. Therefore, in the future work, we will stuy how to evelop efficient approximate algorithms to solve this mixe integer programming problem when the network scale gets larger, such as the branch-an-boun an cutting plane methos. V. NUMERICAL RESULTS In this section, numerical results are provie to emonstrate the performance of our propose scheme. In our simulation, the path-loss exponent of wireless channels is β =3 an the aitive white Gaussian noise power is σ 2 =mw. In aition, we set the total transmit power constraint, P 0 = 000 mw. The 2D square topology of the scenario is shown as in Fig. 2, where there is a source, a estination, an eavesropper an several ranomly istribute caniate noes whose center is at (0, 0). To emonstrate the performance of our propose scheme, two cooperative schemes, ecoe-an-forwar (DF) scheme an cooperative jamming (CJ) scheme mentione in [2] have been involve. The relays transmit a weighte version of a reencoe noise-free message signal for the DF scheme while the relays transmit a jamming signal for the CJ scheme. Accoring to our assumption that all caniate noes can properly ecoe the signals from the source, we can assign a fixe large value of P s = 200 mw to satisfy the assumption, an thus our establishe moel is applie to solve the power allocation an weight factors for the relays an jammers. In aition, the number of caniate noes is N =6. As in Fig. 2, the eavesropper location is first fixe at (0, 5). When move the position of the eavesropper from (5, 5) to (5, 5), we can obtain the relationship between the secrecy rate of the whole transmission an the istance from the eavesropper to the source. From Fig. 3, when the istance between the eavesropper an the source becomes larger, the secrecy rates for the DF scheme, the CJ scheme an our propose scheme first increase an then ecrease. The secrecy rates for all three schemes increase since the receive signal power at the eavesropper ecreases when the eavesropper moves away from the source, while when the istance between the eavesropper an the source continues to increase, the secrecy rate ecreases since the eavesropper approaches the estination, which has a negative impact on the secrecy rate. Furthermore, the secrecy rate for the CJ scheme is much lower than the other two schemes, since the CJ scheme can only ecrease the rate at the eavesropper but it cannot improve the rate at the estination. Fig. 4 shows the secrecy rate for ifferent numbers of caniate noes. The eavesropper location is fixe at (0, 5). From the figure, with the number of caniate noes increasing, the secrecy rate of the whole transmission for each scheme becomes larger, which inicates that increasing the number of caniate noes improves the secrecy rate. However, when the number of caniate noes continues to increase, the secrecy rates for three schemes increase slowly an are almost saturate, which shows that there is a limitation for improving the secrecy rate through increasing the number of caniate noes. In aition, the secrecy rate for the CJ scheme is also much lower than the other two schemes which is limite by the istance between the source an the estination. Furthermore, the secrecy rate for our scheme has a better performance than the other two schemes since we consier the optimal selection of relays an jammers to simultaneously enhance the rate at the estination an reuce the rate at the eavesropper. VI. CONCLUSION In this paper, a joint cooperative relaying an jamming scheme has been presente to enhance the secrecy rate of a 4452

6 Fig. 4. propose scheme DF scheme CJ scheme Number of Caniate Noes Secrecy rate versus number of caniate noes. wireless channel in the presence of an eavesropper, which is an insie attacker. We have propose to assign some relay noes to act as jammers while the rest continue relaying information to the estination. A mixe integer programming problem for maximum secrecy rate has been formulate. Furthermore, our scheme also consiers some special challenges, such as ealing with both the impact of the insie attacker an the phase an frequency synchronization in istribute beamforming. Numerical results have emonstrate the merits of our scheme in terms of secrecy rate, with comparison to DF an CJ schemes. [0] L. Dong, Z. Han, A.P. Petropulu an H.V. Poor, Cooperative jamming for wireless physical layer security, in 2009 IEEE SSP Workshops, Sept. 2009, pp [] R. Zhang, L. Song, Z. Han an B. Jiao, Physical layer security for twoway untruste relaying with frienly jammers, IEEE Trans. Vehicular Tech., vol. 6, no. 8, pp , Oct [2] L. Dong, Z. Han, A.P. Petropulu an H.V. Poor, Improving wireless physical layer security via cooperating relays, in IEEE Trans. Signal Process., vol. 58, no. 3, pp , Mar [3] H.-M. Wang, M. Luo an Q. Yin, Hybri cooperative relaying an jamming for secure two-way relay networks, in Proc. IEEE GLOBECOM, Dec. 202, pp [4] J. Chen, R. Zhang, L. Song, Z. Han an B. Jiao, Joint relay an jammer selection for secure two-way relay networks, IEEE Trans. Inf. Forensics an Security, vol. 7, no., pp , Feb [5] Hong L., M. McNeal an C. Wei, Secure cooperative MIMO communications uner active compromise noes, 20 IEEE Int l. Pervasive Computing an Commun. Workshops (PERCOM Workshops), Mar 20, pp [6] R. Muumbai, G. Barriac an U. Mahow, On the feasibility of istribute beamforming in wireless networks, IEEE Trans. Wireless Commun., vol. 6, no. 5, pp , May [7] R. Liu an W. Trappe, Securing Wireless Communications at the Physical Layer, Springer, [8] T. S. Rappaport, Wireless Communications Principles an Practice, 2n e. Prentice Hall, ACKNOWLEDGMENT This work was supporte in part by the National Natural Science Founation of China (Grant No ), the Science Technology Innovation Founation for Young Teachers in BUPT (Grant No. 203RC0202), the State Major Science an Technology Special Projects (Grant No. 202ZX ), an the Beijing Higher Eucation Young Elite Teacher Project (Grant No. YETP0442). REFERENCES [] A. D. Wyner, The wire-tap channel, Bell Syst. Tech. J., vol. 54, no. 8, pp , Oct [2] S. K. Leung-Yan-Cheong an M. E. Hellman, The Gaussian wire-tap channel, IEEE Trans. Inf. Theory, vol. 24, no. 4, pp , Jul [3] I. Csiszar an J. Korner, Broacast channels with confiential messages, IEEE Trans. Inf. Theory, vol. 24, no. 3, pp , May 978. [4] Q. Guan, F.R. Yu an S. Jiang, Capacity-optimize topology control for MANETs with cooperative communications, IEEE Trans. Wireless Commun., vol. 0, no. 7, pp , Jul. 20. [5] D.W.K. Ng, E.S. Lo an R. Schober, Energy-efficient resource allocation in multi-cell OFDMA systems with limite backhaul capacity, IEEE Trans. Wireless Commun., vol., no. 0, pp , Oct [6] J. Kim, A. Ikhlef an R. Schober, Combine relay selection an cooperative beamforming for physical layer security, J. Commun. Netw., vol. 4, no. 4, pp , Aug [7] Y. Liu an W. Nie, Selection base cooperative beamforming an power allocation for relay networks, J. Commun. Netw., vol. 3, no. 4, pp , Aug. 20. [8] Y. Yang, Q. Li, W.-K. Ma, J. Ge an P. C. Ching, Cooperative secure beamforming for AF relay networks with multiple eavesroppers, IEEE Signal Process. Lett., vol. 20, no., pp , Jan [9] J. Yang, I. Kim an D. I. Kim, Optimal cooperative jamming for multiuser broacast channel with multiple eavesroppers, IEEE Trans. Wireless Commun., vol. 2, no. 6, pp , Jun