Secure Cognitive Reactive Decode-and-Forward Relay Networks: With and Without Eavesdropper
Transcript of Secure Cognitive Reactive Decode-and-Forward Relay Networks: With and Without Eavesdropper
Secure Cognitive Reactive Decode-and-Forward RelayNetworks: With and Without Eavesdropper
Dac-Binh Ha1 • Tung Thanh Vu1 •
Tran Trung Duy2 • Vo Nguyen Quoc Bao2
� Springer Science+Business Media New York 2015
Abstract In this paper, we study performances of cooperative relay networks in cognitive
radio with reactive multiple decode-and-forward (DF) relays. In particular, we first derive
exact and asymptotic expressions of outage probability for the considered scheme with K-
th best relay selection over Rayleigh fading channel. Next, in the presence of an eaves-
dropper and the same relay selection scheme, secrecy performance of the cognitive relay
network is also evaluated, in terms of average secrecy capacity. We then perform Monte-
Carlo simulations to verify the theoretical derivations. Our results have presented the
significance of using relay networks to enhance the system and secrecy performance of
cognitive reactive DF relay networks.
Keywords Underlay cognitive radio � Reactive relay selection � Physical-layer security �Channel capacity � Secrecy capacity
This paper has been presented in part as the best paper in Signal Processing and CommunicationsSymposium at the IEEE CHINACOM, Maoming, China, August 2014.
& Tran Trung [email protected]
Dac-Binh [email protected]
Tung Thanh [email protected]
Vo Nguyen Quoc [email protected]
1 Duy Tan University, Da Nang, Vietnam
2 Posts and Telecommunications Institute of Technology, Hochiminh city, Vietnam
123
Wireless Pers CommunDOI 10.1007/s11277-015-2924-y
1 Introduction
With the ever increasing demand of mobile multimedia services, wireless systems have
encountered several practical constraints, e.g., bandwidth availability, multi-path signal
degradation, interference management. Among these affects, shortage of radio frequency
spectrum is the most critical issue. To cope with this problem, cognitive radio (CR) has
been proposed as an efficient sharing scheme among licensed and unlicensed users. By
allowing unlicensed users to concurrently occupy the radio spectrum, the utilization of
frequency spectrum is remarkably enhanced [1]. A key characteristic of spectrum sharing
networks is to guarantee the quality of service (QoS) of the licensed networks under an
acceptable interference constraint. To accomplish this, unlicensed users need to respect the
interference limit regulated by the licensed users. However, due to the limited transmit
power, the performance at the unlicensed receivers is drastically reduced, especially when
the channels experience heavy path-loss and severe shadowing effects [2, 3].
However, the CR has faced some practical challenges due to its deduced system per-
formance and secure performance as compared to non spectrum-sharing counterpart.
Recently, opportunistic relaying strategies which are based on equal-power multiple-relay
transmissions with local channel knowledge have been realized as a supreme mean to
enhance communication coverage [4]. As such, the extension of using relay in CR has
attracted great attention in the research community [5–8]. In [9], the authors evaluated the
performance of proactive relay networks, in terms of outage probability, error probability,
and ergodic capacity. In [10], the ergodic capacity of reactive multiple decode-and-forward
(DF) relays has been investigated. It is important to note that the published works [9, 10]
assumed that the system can select the best relay among the potential ones to assist the
source–destination communication. However, such scheduling may be not applicable in the
dense heterogeneous networks due to the load balance and imperfect channel state infor-
mation. Very recently, the authors in [11] proposed an incremental cooperative commu-
nication protocol with N-th best partial relay selection. In [12], by considering the N-th best
relay, the outage probability of reactive DF relays has been derived. However, this work is
limited to the case of identical fading channels and only exact outage probability.
Due to the broadcast nature in wireless network, the message can be overheard by
eavesdroppers. To protect transmission against severe privacy security risks from eaves-
droppers, Physical Layer Security has been considered as an attractive approach without
complex cryptographic deployments. In [13], Wyner first introduced the wiretap channel to
study the secrecy rate. Latter researches on the secrecy performance over different fading
channels were extended and analyzed [14–17]. Specifically, in [14, 15], the authors have
studied the main and eavesdropper links undergo independent Rayleigh fading and eval-
uated the secrecy capacity. More recently, [16] evaluated the secrecy performance over
diffused power fading channels while [17] investigated multiple antennas wiretap channels
in cognitive network.
To improve the secrecy performance of wireless network, antenna selection techniques
in multiple-input-multiple-output (MIMO) systems [18–20] have been studied to enhance
the secrecy performance. Besides antenna selection technique, relaying technique can be
used to improve the secrecy performance of wireless networks. For example, ergodic
secrecy capacity and the impact of large scale antenna arrays at either the source or the
destination were investigated in [21]. In [22], the authors have considered the security
enhancement in cooperative single carrier systems with multiple relays, multiple desti-
nations and multiple eavesdroppers in selective fading channels. For security improvement
D.-B. Ha et al.
123
of cognitive radio network, the relaying technique has been also exploited. In particular,
[23] introduced a selected pair of opportunistic cognitive relays to protect secure trans-
mission against eavesdroppers. In this paper, our contributions are summarized as follow:
• We consider the joint impact of peak interference power constraint of licensed user and
maximal transmit power of unlicensed user on the performance of CR networks with
reactive multiple DF relays and K-th best relay selections.
• We derive exact outage probability and ergodic capacity for the proposed protocol
without eavesdropper. Moreover, we consider a more general fading model in which all
links are assumed independent but non-identically distributed (i.n.i.d.).
• To provide more insights into the system performance, we also derive the asymptotic
outage probability where both diversity and coding gains are obtained. It has been
shown that the full diversity can be realized when the peak interference power is
proportional to the maximal transmit power.
• Finally, we also derive asymptotic secrecy capacity of the cooperative hop in the
presence of an eavesdropper when the primary user is far the secondary network. It has
been showed that secrecy performance is enhanced by increasing the number of relays.
2 System and Channel Models
The cognitive network consists of a secondary source (S), a secondary destination (D), M
secondary relays (R), a primary user (PU) and an eavesdropper (E). Let dsi; did; dsp; dip and
die denote distances of the S ! Ri, Ri ! D, S ! PU, Ri ! PU and Ri ! E links,
respectively, where i 2 1; 2; . . .;Mf g. We also denote hsi; hid; hsp; hip, and hie as channel
coefficients of the S ! Ri, Ri ! D, S ! PU, Ri ! PU, and Ri ! E links, respectively.
We assume that all of the channels follow a Rayleigh fading distribution. Hence, the
channel gains csi ¼ jhsij2; cid ¼ jhidj2; csp ¼ jhspj2, cip ¼ jhipj2, and cie ¼ jhiej2 follow
exponential distributions. To take path-loss into account, we model the parameters of
csi; cid; csp, cip, and cie as in [24]: ksi ¼ ðdsiÞb, kid ¼ ðdidÞb, ksp ¼ ðdspÞb, kip ¼ ðdipÞb, andkie ¼ ðdieÞb, where b denotes the path-loss exponent.
In cognitive underlay networks, the source and relay must adapt their transmit power so
that interference caused at PU is lower than a maximum interference level, denoted by Ith.
In addition, it is also assumed that their transmit power must be lower than a maximum
threshold, denoted by Pth. We assume that all of the nodes are equipped with a single
antenna and operate on half-duplex mode. Similar to the model proposed in [6], the
maximum transmit power of the source S and relay Ri are, respectively, given as PS ¼min Pth; Ith=csp
� �and PRi
¼ min Pth; Ith=cip� �
. Therefore, the instantaneous signal-to-noise
ratio (SNR) of the S ! Ri and Ri ! D links are, respectively, expressed as follow
Wsi ¼Mmin Pth; Ith=csp� �
N0
csi ¼ min �cP;�cIcsp
!
csi;
Wid ¼Mmin Pth; Ith=cip
� �
N0
cid ¼ min �cP;�cIcip
!
cid;
ð1Þ
where N0 denotes the variance of an additive complex Gaussian noise (it is assumed to be
the same at all receivers in the relays destination and eavesdropper). Also, we define
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
quantities as follows: �cP ¼ Pth=N0 and �cI ¼ Ith=N0. Without loss of generality, we assume
that the ratio between �cP and cI is constant, i.e.,
�cI�cP
¼ Ith
Pth
¼ l: ð2Þ
The operation of the proposed protocol is realized by TDMA technique. In the first time
slot, the source S broadcasts its data to the relays. Then, the relays try to decode the
source’s signal from the received signal. Let us denote Q1 and Q2 as the set of the relays
decoding the signal successfully and unsuccessfully, respectively. We can assume that
Q1 ¼ Rj1 ;Rj2 ; . . .;RjN
� �and Q2 ¼ RjNþ1
;RjNþ2; . . .;RjM
� �, where N is the cardinality of Q1,
N 2 0; 1; 2; . . .;Mf g, and j1; j2; . . .; jM 2 1; 2; . . .;Mf g. In the considered cognitive radio
networks, we consider two cases as follow:
• (C1): The Kth-best relay is chosen among N successful relays (N�K) to forward the
source’s signal to the destination at the second time slot. The relay selection is realized
by the following strategy
Rb:Wbd ¼ Kth maxt¼1;2;...;N
Wjtd
� �; ð3Þ
where Rb denotes the Kth-best chosen relay.
• (C2): The system cannot choose a relay to forward the source’s signal to the destination
since N\K. Hence, in this case, the signal is dropped.
3 Performance Evaluation
3.1 System Performance of Cognitive Relay Networks Without Eavesdropper
3.1.1 Derivation of the CDF and PDF of the Instantaneous SNR by Kth-Best RelaySelection
We observe that among N successful relays of the set Q1, there are K � 1 relays whoseWjtd
is larger than Wbd and N � K relays whose Wjtd is smaller than Wbd, where
t 2 1; 2; . . .;Nf g= cjjc ¼ bf gf g. We, respectively, denote these sets as W1 ¼Rz1 ;Rz2 ; . . .;RzK�1f g and W2 ¼ RzKþ1
;RzKþ2; . . .;RzN
� �, where W1 � Q1,W2 � Q1 and
W1 [W2 ¼ Q1= Rbf g. Thus, from [25, Eq. (8)], we can write the CDF and PDF of
Wid; i 2 1; 2; . . .;Mf g, as follow
FWidðxÞ ¼ 1� exp � kid
�cPx
� �þ kidxkidxþ kipl�cP
exp � kid�cP
x� kipl
� �;
fWidðxÞ ¼ kid
�cPexp � kid
�cPx
� �exp � kid
�cPx� kipl
� �
� kidkipl�cPkidxþ kipl�cP� �2 �
k2idx
�cP kidxþ kipl�cP� �
" #
:
ð4Þ
Let us denote Y1 ¼ minðWz1d; . . .;WzK�1dÞ and Y2 ¼ maxðWzKþ1d; . . .;WzNdÞ, we can express
the CDFs of the RVs Y1 and Y2 as
D.-B. Ha et al.
123
FY1ðxÞ ¼ 1�YK�1
v¼1
exp � kzvd�cP
x
� �� kzvdxkzvdxþ kzvpl�cP
exp � kzvd�cP
x� kzvpl
� �� �;
FY2ðxÞ ¼YM
v¼Kþ1
1� exp � kzvd�cP
x
� �þ kzvdxkzvdxþ kzvpl�cP
exp � kzvd�cP
x� kzvpl
� �� �:
ð5Þ
In addition, we can formulate the CDF of Wbd as follow
FWbdðxÞ ¼
X
c¼1
N
jc¼b
X
W1;W2
PrðWbd\x; Y2 �Wbd � Y1Þ; ð6Þ
which is equivalent to the following expression
FWbdðxÞ ¼
X
c¼1
N
jc¼b
Z x
0
kbd expð�kbdy=�cPÞ�cP
þ kbdkbpl�cPðkbdyþ kbpl�cPÞ2
� ðkbdÞ2y�cP kbdyþ kbpl�cP� �
!"
� exp � kbd�cP
y� kbpl
� ��ð1� FY1ðyÞÞFY2ðyÞdy: ð7Þ
Based on (7), the corresponding PDF is given by
fWbdxð Þ ¼ oFWbd
xð Þox
; ð8Þ
which is evaluated as
fWbdðxÞ ¼
X
c¼1
N
jc¼b
X
W1;W2
kbd exp �kbdx=�cPð Þ�cP
þ kbdkbpl�cPkbdxþ kbpl�cP� �2 �
ðkbdÞ2x�cP kbdxþ kbpl�cP� �
!"
� exp � kbd�cP
x� kbpl
� ��ð1� FY1ðxÞÞFY2ðxÞ: ð9Þ
Substituting (5) into (9), the PDF of Wbd is given as in (10).
fWbdðxÞ ¼
X
c¼1
N
jc¼b
X
W1;W2
kbd�cP
exp �kbd�cP
x
� �þkbdkbpl�cP exp � kbd
�cPx� kbpl
kbdxþ kbpl�cP� �2
2
4
�kbd�cP
kbdxexp � kbd�cPx� kbpl
kbdxþ kbpl�cP
3
5YK�1
v¼1
exp �kzvd�cP
x
� ��kzvdxexp � kzvd
�cPx� kzvpl
kzvpxþ kzvpl�cP
2
4
3
5
�YK�1
v¼1
exp �kzvd�cP
x
� ��kzvdxexp � kzvd
�cPx� kzvpl
kzvdxþ kzvpl�cP
2
4
3
5: ð10Þ
3.1.2 Exact Outage Probability of Non-homogeneous Networks
We assume that the relay Ri (the destination D) can decode the signal successfully if the
instantaneous SNR of the S ! Ri (Ri ! D) link exceeds a threshold cth. Therefore, theoutage probability of the proposed protocol can be calculated as follow
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
Pout ¼ Pr N\Kð Þ|fflfflfflfflfflffl{zfflfflfflfflfflffl}
Pout1
þPr Wbd\cth;N �Kð Þ|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Pout2
; ð11Þ
where Pout1 represents case C1 where the system cannot choose any relays to forward the
source’s signal to the destination, while Pout2 represents case C2 where the system can
choose the Kth-best relay for the cooperation but the transmission between the selected
relay and the destination is in outage. Considering the outage probability Pout1 , we can
formulate it as
Pout1 ¼
X
Q1;Q2
N\K
Pr Wsj1 � cth; . . .;WsjN � cth;WsjNþ1\cth; . . .;WsjM\cth
� �; ð12Þ
which can be rewritten by (13) at the top of next page,
Pout1 ¼
X
Q1;Q2
N\K
Pr csp\l; csj1 �qP; . . .; csjN �qP; csjNþ1\qP; . . .; csjM\qP
� �
þ Pr csp � l; csj1 �qIcsp; . . .; csjN �qIcsp; csjNþ1\qIcsp; . . .; csjM\qIcsp
� �
¼MX
Q1;Q2
N\K
ðV1 þ V2Þ;ð13Þ
where
V1 ¼M Pr csp\l; csj1 � qP; . . .; csjN �qP; csjNþ1\qP; . . .; csjM\qP
� �;
V2 ¼M Pr csp � l; csj1 �qIcsp; . . .; csjN �qIcsp; csjNþ1\qIcsp; csjM\qIcsp
� �;
with qP ¼ cth=cP and qI ¼ cth= lcPð Þ. With some manipulations, we can readily obtain V1
and V2 as
V1 ¼ 1� exp �kspl� �� �YN
t¼1
exp �ksjtqP� � YM
t¼Nþ1
1� exp �ksjtqP� �� �
; ð14Þ
V2 ¼Z þ1
lksp expð�kspxÞ
YN
t¼1
expð�ksjtqIxÞYM
t¼Nþ1
ð1� exp �ksjtqIx� �
Þ" #
dx: ð15Þ
Having expandedQM
t¼Nþ1 1� exp �ksjtqIx� �� �
by the binomial identity, we have the
following expression for V2
V2 ¼ksp exp �kspl�
PNt¼1 ksjtqP
� �
ksp þPN
t¼1 ksjtqIþXM�N
v¼1
X
j1;...;jv¼Nþ1
M
j1\...\jv
ð�1ÞvkspqPksp þ
PNt¼1 ksjt þ
Pvl¼1 ksjl
� �qI
� exp �kspl�XN
t¼1
ksjt þXv
l¼1
ksjl
!cthcP
!
: ð16Þ
Collecting (14) and (16), the closed-form expression of the exact outage probability (13) is
given by
D.-B. Ha et al.
123
Pout1 ¼
X
Q1;Q2
N\K
ð1� expð�ksplÞÞYN
t¼1
exp �ksjtcth�cP
� � YM
t¼Nþ1
1� exp �ksjtcth�cP
� �� �"
þkspl�cP exp �kspl�
PNt¼1 ksjt
cth�cP
kspl�cP þPN
t¼1 ksjtcth
þXM�N
v¼1
X
j1;...;jv¼Nþ1
M
j1\...\jv
ð�1Þvkspl�cPkspl�cP þ
PNt¼1 ksjt þ
Pvl¼1 ksjl
� �cth
� exp �kspl�XN
t¼1
ksjt þXv
l¼1
ksjl
!cth�cP
!#
:
ð17Þ
Note that in the derivation of (17), we used qP ¼M
cth=�cP and qI ¼M
cth= l�cPð Þ.Next, we calculate the term Pout
2 in (11). We first rewrite it as follow
Pout2 ¼
X
Q1;Q2
N�K
PrðWsj1 � cth; . . .;WsjN � cth;WsjNþ1\cth; . . .;WsjM\cthÞFWbd
ðcthÞ: ð18Þ
With the same manner as in the derivation of (17) and using (7), we can obtain the exact
expression of Pout2 as in (19).
Pout2 ¼
X
Q1;Q2
N�K
1�expð�ksplÞ� �YN
t¼1
exp �ksjtcth�cP
� � YM
t¼Nþ1
1�exp �ksjtcth�cP
� �� �"
þ kspl�cPkspl�cPþ
PNt¼1ksjtcth
exp �kspl�XN
t¼1
ksjtcth�cP
!
þXM�N
v¼1
X
j1;...;jv¼Nþ1
M
j1\...\jv
ð�1Þvkspl�cP exp �kspl�PN
t¼1ksjt þPv
l¼1ksjl� �
cth=�cP� �
kspl�cPþPN
t¼1ksjt þPv
l¼1ksjl� �
cth
#
X
c¼1
N
jt¼b
X
W1;W2
Z cth
0
kbd�cP
exp �kbd�cP
y
� �þkbdkbpl�cP expð�kbdy=�cP�kbplÞ
ðkbdyþkbpl�cPÞ
2"
�kbd�cP
kbdyexp �kbdy=�cP�kbpl� �
kbdyþkbpl�cP
#YK�1
v¼1
exp �kzvd�cP
y
� ��kzvdyexp �kzvdy=�cP�kzvpl
� �
kzvdyþkzvpl�cP
� �
YN
v¼Kþ1
1�exp �kzvd�cP
y
� �þkzvdyexp �kzvdy=�cP�kzvpl
� �
kzvdyþkzvpl�cP
� �!
dy:
ð19Þ
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
3.1.3 Exact Outage Probability of Homogeneous Networks
In these networks, we assume that ksi ¼ ks, kid ¼ kd, and kip ¼ kp for all i, so that (13) can
be rewritten as follow
PH;out1 ¼
XK�1
N¼0
M
N
� �ðVH
1 þ VH2 Þ; ð20Þ
where
VH1 ¼ 1� exp �kspl
� �� �exp �NksqPð Þ 1� exp �ksqPð Þð ÞM�N ;
and
VH2 ¼
Z þ1
lksp expð�kspxÞ expð�NksqIxÞð1� exp �ksqIxð ÞÞM�N �
dx: ð21Þ
Similar to the derivation of Pout1 in non-homogeneous networks, P
H;out1 can be obtained as
PH;out1 ¼
XK�1
N¼0
M
N
� �ð1� expð�ksplÞÞ exp
�Nkscth�cP
� �1� exp
�kscth�cP
� �� �M�N"
þXM�N
t¼0
M � N
t
� �kspl�cP
kspl�cP þ ðN þ tÞkscthexp �kspl� ðN þ tÞkscth
�cP
� �#
:
ð22Þ
In these networks, by using the K-th best order statistics, Pr W2jc\cth� �
is given by
FWbdðxÞ ¼
XK
t¼1
N
t � 1
� �ðFWzt d
ðxÞÞN�tþ1ð1� FWzt dðxÞÞt�1: ð23Þ
By using the results obtained in (4) and (23), we can express PH;out2 by (24).
PH;out2 ¼
XM
N¼K
M
N
� �"
ð1� expð�ksplÞÞ expð�Nkscth=�cPÞð1� expð�kscth=�cPÞÞM�N :
þXM�N
t¼0
�1ð Þt M � N
t
� �kspl�cP
kspl�cP þ N þ tð Þkscthexp � ksplþ N þ tð Þkscth=�cP
� �� �#
XK
t¼1
N
t � 1
� �1� expð�kd
cth�cPÞ þ kdcth
kdcth þ kpl�cPexpð�kp
cth�cP
� kplÞ� �N�tþ1
exp �kdcth�cP
� �þ kdcthkdcth þ kpl�cP
exp �kdcth�cP
� kpl
� �� �t�1
:
ð24Þ
Finally, by adding either Pout1 and Pout
2 or PH;out1 and P
H;out2 , we obtain a closed-form
expression for the outage probability. To see an asymptotic outage diversity gain, we will
make an asymptotic outage probability analysis next.
D.-B. Ha et al.
123
3.1.4 Asymptotic Outage Probability
We shall derive the expressions of Pout in the high qP region. By choosing only first two
terms of Maclaurin expansion series for g(x) a function of x, we have
gðxÞ x!0�g 0ð Þ þ og
oxjx¼0
x. Applying this result for expð�xÞ, we obtain expð�xÞ x!0
�1� x
and 1� expð�xÞ x!0�1. Thus, at a very high �cP value, i.e., �cP ! þ1 (or qP ! 0), we can,
respectively, obtain asymptotic (25) and (26) from (14) and (15) as follows
V1 �qp!0
1� exp �kspl� �� � YM
t¼Nþ1
ksjt
!
qM�NP ; ð25Þ
V2 �qp!0
kspYM
t¼Nþ1
ksjt
!qPl
� �M�NZ þ1
lxM�N exp �kspx
� �dx
¼YM
t¼Nþ1
ksjt
!
C M � N þ 1; kspl� � qP
kspl
� �M�N
;
ð26Þ
where Cða; xÞ ¼Rþ1x
xa�1 exp �xð Þdx denotes the incomplete gamma function [26]. Sim-
ilarly, an asymptotic CDF of Wid is given by
FWidðxÞ �
qp!0
kid þkid expð�kipuÞ
kipl
� �x
�cP; ð27Þ
which results in the asymptotic PDF of Wid in the following form
fWidðxÞ �
qp!0
kid þkid exp �kipu
� �
kipl
� �1
�cP: ð28Þ
Now using Eqs. (27) and (28), we can first obtain
Pr Wbd\cthð Þ �qp!0X
c¼1
N
jc¼b
X
W1;W2
kbd þkbd expð�kbpuÞ
kbpl
� �1
�cP
� �N�Kþ1
�YN
v¼Kþ1
kzvd þkzvd exp �kzvpl
� �
kzvpu
� �Z cth
0
xN�Kdx;
ð29Þ
which becomes
PrðWbd\cthÞ �qp!0X
c¼1
N
jc¼b
kbd þkbd expð�kbplÞ
kbpl
� �
�X
W1;W2
YN
v¼Kþ1
kzvd þkzvd expð�kzvplÞ
kzvpl
� �qN�Kþ1P
N � K þ 1:
ð30Þ
Theorem 1 Using Eqs. (25) and (26), and (30), asymptotic outage probabilities for non-
homogeneous and homogeneous networks are, respectively, given by
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
~Pout ��cp!1
ðD1 þ D2Þcth�cP
� �M�Kþ1
;
~PH;out ��cp!1
ðD3 þ D4Þcth�cP
� �M�Kþ1
;
ð31Þ
where we define
D1 ¼MX
Q1;Q2
N\K
XM
t¼K
ksjt
!
qPð ÞM�Kþ11� expð�ksplÞ þ
CðM � K þ 2; ksplÞðksplÞM�Kþ1
!
;
D2 ¼MX
Q1;Q2
N �K
XM
t¼Nþ1
ksjt
!
qPð ÞM�Kþ11� expð�ksplÞ þ
C M � N þ 1; kspl� �
kspl� �M�N
!
�X
c¼1
N
jc¼b
kbd þkbd exp �kbpuð Þ
kbpu
� �
N � K þ 1
X
W1;W2
YN
v¼Kþ1
kzvd þkzvd expð�kzvpuÞ
kzvpl
� �;
D3 ¼MM
K � 1
� �� 1� exp �kspl
� �� �ksð ÞM�Kþ1 þ CðM � K þ 2; ksplÞ ks=kspl
� �M�Kþ1h i
;
D4 ¼MXM
N¼k
ð1� expð�ksplÞÞ ksð ÞM�N þ CðM � N þ 1; ksplÞ ks=kspl� �M�N
h i:
From (31), we can see that the outage diversity gain is Gd ¼ M � K þ 1.
Proof A proof of this theorem is provided in ‘‘Appendix’’.
Note that from Theorem 1, Gd is in the range of [1, M].
3.1.5 Ergodic Channel Capacity
The channel capacity of the proposed protocol can be expressed as
C Wbdð Þ ¼0; if N\K;
1
2log2ð1þWbdÞ; if N�K:
(
ð32Þ
From (32), the average channel capacity can be formulated as
Cavg ¼1
2 logð2ÞX
Q1;Q2
N�K
PrðWsj1 � cth; . . .;WsjN � cth;WsjNþ1\cth; . . .;WsjM\cthÞ
�Z 1
0
logð1þ xÞfWbdðxÞdx:
ð33Þ
Combining results obtained in (10) and (19), the exact expression of (33) is given by (34)
(see the top of next page).
D.-B. Ha et al.
123
Cavg ¼1
2 logð2ÞX
Q1;Q2
N�K
1� expð�ksplÞ� �YN
t¼1
exp �ksjtcth�cP
� � YM
t¼Nþ1
1� exp �ksjtcth�cP
� �� �"
þ kspl�cPkspl�cP þ
PNt¼1 ksjtcth
exp �kspl�XN
t¼1
ksjtcth�cP
!
þXM�N
v¼1
X
j1;...;jv¼Nþ1
M
j1\...\jv
ð�1Þvkspl�cP exp �kspl�PN
t¼1 ksjt þPv
l¼1 ksjl� �
cth=�cP� �
kspl�cP þPN
t¼1 ksjt þPv
l¼1 ksjl� �
cth
#
�X
c¼1
N
jt¼b
X
W1;W2
Z 1
0
logð1þ xÞ kbd�cP
exp � kbd�cP
x
� �þkbdkbpl�cP exp � kbd
�cPx� kbpl
kbdxþ kbpl�cP� �2
2
4
� kbd�cP
kbdx exp � kbd�cPx� kbpl
kbdxþ kbpl�cP
3
5YK�1
v¼1
exp � kzvd�cP
x
� ��kzvdx exp � kzvd
�cPx� kzvdl
kzvdxþ kzvdl�cP
2
4
3
5
YK�1
v¼1
exp � kzvd�cP
x
� ��kzvdx exp � kzvd
�cPx� kzvpl
kzvdxþ kzvpl�cP
2
4
3
5dx:
ð34Þ
3.2 Secrecy Performance of Cognitive Relay Network with Eavesdropper
In the considered network, we assume that an eavesdropper E appears and attempts to
overhear the data transmitted from the relays. We assume that the eavesdropper is far the
source and hence it cannot listen the data transmission from this node.
Similar as above, we can express the channel capacity of the data link between the
selected relay Rb to the destination D as
Cbd ¼1
2log2ð1þWbdÞ: ð35Þ
For the eavesdropper link, the channel capacity from Rb to E can be given as
Cbe ¼1
2log2ð1þWbeÞ; ð36Þ
where wbe ¼ min cP; cI=cbp� �
cbe.Combining (35) and (36), the secrecy capacity can be defined as [27]:
CSec ¼ maxðCbd � Cbe; 0Þ
¼ max1
2log2
1þWbd
1þWbe
� �; 0
� �:
ð37Þ
Moreover, as discussed above, the considered system can select a relay for the cooperation
as Pr N�Kð Þ. Hence, the average secrecy capacity can be given as
CavgSec ¼ Pr N�Kð ÞE CSecf g; ð38Þ
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
where Eð:Þ is the expected operator.
Since it is difficult to obtain the exact expression of the average secrecy rate CavgSec , we
attempt to obtain an asymptotic one in case that the primary user is far the secondary
network. Indeed, in this case, we have the approximations as follows: wbd � cPcbd and
wbe � cPcbe, and the secrecy capacity in (37) can be approximated by
CSec � max1
2 log 2ð Þ log1þ cPcbd1þ cPcbe
� �; 0
� �: ð39Þ
Before calculating the average secrecy rate, we must find the CDF of the random variable
Z, i.e., Z ¼ 1þcPcbd1þcPcbe
. Indeed, the CDF FZ zð Þ, z� 1, can be formulated as
FZ zð Þ ¼ Pr Z\zð Þ ¼ Pr cbd\z� 1
cPþ zcbe
� �
¼Z þ1
0
Fbd
z� 1
cPþ zy
� �fcbe yð Þdy:
ð40Þ
Here, we again consider two cases: non-homogeneous network and homogeneous network.
3.2.1 Non-homogeneous Networks
In non-homogeneous networks, the CDF FbdðxÞ in (40) can be calculated similarly to (7) as
Fbd xð Þ ¼X
c¼1
N
jc¼b
Z x
0
kbd exp �kbdyð Þ 1� FT1 yð Þð ÞFT2 yð Þdy; ð41Þ
where T1 ¼ min cz1d; . . .; czK�1d
� �, T2 ¼ max czKþ1d
; . . .; czNd� �
, and their CDFs can be
respectively expressed by
FT1 yð Þ ¼ 1� exp �XK�1
v¼1
kzvdy
!
;
FT2 yð Þ ¼YM
v¼Kþ1
1� exp �kzvdyð Þð Þ
¼ 1þXM�K
j¼1
�1ð ÞjX
i1¼...¼ij¼Kþ1
M
i1 [ ...[ ij
exp �Xj
k¼1
kzik dy
!
:
ð42Þ
Substituting (42) into (41), and after some manipulation, we can obtain
Fbd xð Þ ¼X
c¼1
N
jc¼b
kbdx1
1� exp �x1xð Þð Þ
þX
c¼1
N
jc¼b
XM�K
j¼1
�1ð ÞjX
i1¼...¼ij¼Kþ1
M
i1 [ ...[ ij
kbdx2
1� exp �x2xð Þð Þ;ð43Þ
where x1 ¼ kbd þPK�1
v¼1
kzvd and x2 ¼ x1 þPj
k¼1
kzik d .
D.-B. Ha et al.
123
Plugging (40), (43) and fcbe yð Þ ¼ kbe exp �kbeð Þ together, the CDF FZ zð Þ can be given as
FZ zð Þ ¼X
c¼1
N
jc¼b
kbdx1
� kbdx1
:kbe exp x1=cPð Þ
kbe þ x1zexp �x1z
cP
� �� �
þX
c¼1
N
jc¼b
XM�K
j¼1
�1ð ÞjX
i1¼...¼ij¼Kþ1
M
i1 [ ...[ ij
kbdx2
� kbdx2
:kbe exp x2=cPð Þ
kbe þ x1zexp �x2z
cP
� �� �:
ð44Þ
Moreover, the average secrecy capacity can be expressed under the following expression:
CavgSec �
Pr N�Kð Þ2 log 2ð Þ
Z þ1
1
log zð ÞfZ zð Þdz
� Pr N�Kð Þ2 log 2ð Þ
Z þ1
1
1� FZ zð Þz
dz:
ð45Þ
Substituting (44) into (45), we can rewrite CavgSec as follow
CavgSec �
Pr N �Kð Þ2 log 2ð Þ
�
P
c¼1
N
jc¼b
x1 exp kbd=x1ð Þkbd
Z þ1
1
1
z� 1
kbex1=kbd þ z
� �exp � kbd
x1
z
� �dz
þP
c¼1
N
jc¼b
PM�K
j¼1
�1ð ÞjP
i1¼...¼ij¼Kþ1
M
i1 [ ...[ ij
x2 exp kbd=x2ð Þkbd
�Rþ11
1
z� 1
kbex2=kbd þ z
� �exp � kbd
x2
z
� �dz
8>>>>>>>>>>><
>>>>>>>>>>>:
9>>>>>>>>>>>=
>>>>>>>>>>>;
:
ð46Þ
Applying the following integral
Z þ1
1
exp �axð Þbþ x
dx ¼ exp abð ÞE1 1þ bð Það Þ;
for the corresponding integral in (46), we finally obtain
CavgSec � Pr N�Kð Þ
2 log 2ð Þ
�
P
c¼1
N
jc¼b
kbdx1
expx1
cP
� �E1
x1
cP
� �� exp
kbecP
� �E1 1þ kbe
x1
� �x1
cP
� �� �� �
þP
c¼1
N
jc¼b
PM�K
j¼1
�1ð ÞjP
i1¼...¼ij¼Kþ1
M
i1 [ ...[ ij
kbdx2
expx2
cP
� �E1
x2
cP
� �� exp
kbecP
� �E1 1þ kbe
x2
� �x2
cP
� �� �� �
8>>>>>>>>>>><
>>>>>>>>>>>:
9>>>>>>>>>>>=
>>>>>>>>>>>;
;
ð47Þ
where a and b are positive numbers, and E1ð:Þ is the exponential integral function [26].
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
Then, by using the results Pr N�Kð Þ in (19), we can obtain an asymptotic expression of
CavgSec when the primary user is far the secondary system, as presented in (48).
CavgSec �
1
2 log 2ð ÞX
Q1;Q2
N�K
1� expð�ksplÞ� �YN
t¼1
exp �ksjtcth�cP
� � YM
t¼Nþ1
1� exp �ksjtcth�cP
� �� �"
þ kspl�cP
kspl�cP þPN
t¼1
ksjtcth
exp �kspl�XN
t¼1
ksjtcth�cP
!
þXM�N
v¼1
X
j1;...;jv¼Nþ1
M
j1\...\jv
ð�1Þvkspl�cP exp �kspl�PN
t¼1 ksjt þPv
l¼1 ksjl� �
cth=�cP� �
kspl�cP þPN
t¼1 ksjt þPv
l¼1 ksjl� �
cth
3
775
�
P
c¼1
N
jc¼b
kbdx1
expx1
cP
� �E1
x1
cP
� �� exp
kbecP
� �E1 1þ kbe
x1
� �x1
cP
� �� �� �
þP
c¼1
N
jc¼b
PM�K
j¼1
�1ð ÞjP
i1¼...¼ij¼Kþ1
M
i1 [ ...[ ij
kbdx2
expx2
cP
� �E1
x2
cP
� �� exp
kbecP
� �E1 1þ kbe
x2
� �x2
cP
� �� �� �
8>>>>>>>>>>><
>>>>>>>>>>>:
9>>>>>>>>>>>=
>>>>>>>>>>>;
:
ð48Þ
3.2.2 Homogeneous Networks
In these networks, with ksi ¼ ks, kid ¼ kd, kip ¼ kp and kie ¼ ke for all i, with the same
manner, the approximate expression of E Csecf g is given as
E Csecf g � 1
2 log2
�
PK
j¼1
PN�Kþ1
i¼0
�1ð ÞiN
j� 1
� �N � jþ 1
i
� �exp iþ j� 1ð Þkd
cP
� �
E1 iþ j� 1ð ÞkdcP
� �� exp
kecP
� �E1 iþ j� 1ð Þ 1þ ke
iþ j� 1ð Þkd
� �kdcP
� �� �
8>>><
>>>:
9>>>=
>>>;
:
ð49Þ
Moreover, from (24), we can obtain the probability Pr N�Kð Þ. Hence, the approximate
expression of the average secrecy capacity in the i.i.d. network can be given as in (50).
D.-B. Ha et al.
123
CavgSec �
1
2log2
XM
N¼K
M
N
� ��ð1� expð�ksplÞÞexpð�Nkscth=�cPÞð1� expð�kscth=�cPÞÞM�N :
þXM�N
t¼0
�1ð Þt M�N
t
� �kspl�cP
kspl�cPþ Nþ tð Þkscthexp � ksplþ Nþ tð Þkscth=�cP
� �� �#
�
PK
j¼1
PN�Kþ1
i¼0
�1ð ÞiN
j�1
� �N� jþ1
i
� �exp iþ j�1ð Þkbd
cP
� �
E1 iþ j�1ð ÞkbdcP
� �� exp
kbecP
� �E1 iþ j�1ð Þ 1þ kbe
iþ j�1ð Þkbd
� �kbdcP
� �� �
8>>><
>>>:
9>>>=
>>>;
:
ð50Þ
3.3 Performance Comparison
In this subsection, we compare the performance of our scheme with that of proactive relay
selection protocols. At first, we consider the amplify-and-forward relaying protocol using
partial relay selection (denoted by AF-PRS), which was proposed in [3]. In the AF-PRS
scheme, the K-th best relay is selected by [3, eq. (1)]:
Rb: csb ¼ Kth maxm¼1;2;...;M
csmð Þ: ð51Þ
By using [3, eq. (6)] and (1), the outage probability of this scheme can be calculated by
PAF�PRSout ¼ Pr
WsbWbd
Wsb þWbd þ 1\cth
� �: ð52Þ
Moreover, the outage probability of the DF protocol using the K-th partial relay selection
above [28] (denoted by DF-PRS) can be given by
PDF�PRSout ¼ Pr min Wsb;Wbdð Þ\cthð Þ: ð53Þ
Next, we introduce the opportunistic relay selection scheme (denoted by DF-ORS), in
which the K-th best relay is selected as [29, eq. (6)]:
Rc: min Wsc;Wcdð Þ ¼ Kth maxm¼1;2;...;M
min Wsm;Wmdð Þð Þ: ð54Þ
Then, the outage probability of the DF-ORS scheme is computed as
PDF�ORSout ¼ Pr min Wsc;Wcdð Þ\cthð Þ: ð55Þ
4 Numerical Results
In this section, we present Monte Carlo simulations to verify our derivations. In a two-
dimensional Oxy, we assume that the co-ordinates of the source, the destination, the relay,
the primary user and the eavesdropper are (0, 0), (1, 0), xRi; 0ð Þ, xP; yPð Þ and xE; yEð Þ,
respectively. Hence, the distances are calculated as dsi ¼ xR, did ¼ 1� xR,
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
dsp ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix2P þ y2P
p, dip ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffixR � xPð Þ2 þ y2P
qand die ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffixR � xEð Þ2 þ y2E
q. In all of the sim-
ulations, we assume the path-loss exponential b equals 3 and the ratio l between Ith and Pth
equals 1.
In Fig. 1, we present the outage probability as a function of �cP in dB. In this simulation,
the number of relays M equals to 3, the co-ordinates xRiof relays are 0.2, 0.4 and 0.6, and
the positions of the primary user are (0.5, 0.5). It can be seen from this figure that the
outage performance is best when the system can select the best relay to help the source
forward the data to the destination. Figure 2 illustrates the outage probability as a function
of �cP in dB in the i.i.d. network with xRi¼ 0:4, xP ¼ yP ¼ 0:4 and K ¼ 1. It can be
observed that the outage performance is better when increasing the number of relays M.
From Figs. 1 and 2, we can see that the simulation results (Ex) match very well with the
theoretical results (An and As), which validates our derivations. Moreover, it is also seen
that the diversity order obtained equals to M � K þ 1.
Figure 3 compares the outage performance of the proposed protocol with that of the
AF-PRS, DF-PRS and the DF-ORS protocols in the i.i.d. network when xRi¼ 0:5,
xP ¼ 0, yP ¼ 0:5, cth ¼ 1:5, M ¼ 4 and K ¼ 2. It is seen that the proposed protocol out-
performs two partial relay selection protocols, i.e., AF-PRS and DF-PRS, because it
obtains higher diversity gain. Moreover, we can observe from this figure that the DF-ORS
protocol provides the best performance, but it has the same diversity slope with our
scheme.
0 2 4 6 8 10 12 14 16 18 2010−6
10−5
10−4
10−3
10−2
10−1
100
101
γP [dB]
Out
age
Pro
babi
lity
Ex [K=3]An [K=3]As [K=3]Ex [K=2]An [K=2]As [K=2]Ex [K=1]An [K=1]As [K=1]
Fig. 1 The outage probability as a function of �cP in dB when xRi2 0:2; 0:4; 0:6f g, xP ¼ yP ¼ 0:5, cth ¼ 1
and M ¼ 3
D.-B. Ha et al.
123
In Fig. 4, we investigate the impact of the position of the primary user on the average
channel capacity. In this figure, we fix the parameters as follows: M ¼ 2, and xRi2
0:4; 0:6f g and K ¼ 1. It can be seen that the performance of the considered scheme
significantly enhances when the primary user is far the secondary network (xP and yP are
high).
In Figs. 5, 6 and 7, we present the average secrecy capacity of the proposed scheme
when there exists the eavesdropper in the network. In Fig. 5, we place the the primary user
at different positions and observe the impact of the primary user’s position on the secrecy
performance. In particular, we change the values of xP and yP (xP ¼ yP) from 0.2 to 5,
while fixing the parameters M, cth, yE, xE and cP by 4, 0.75, 1, 0.5 and 5 dB, respectively.
Similar to the results presented in Fig. 1, the secrecy performance of the proposed scheme
is better when the system can select the better relay for the cooperation. Furthermore, the
secrecy capacity increases with the increasing of xP and yP. However, the secrecy capacity
converges to a positive constant at the high xP (yP) value. It is because that the impact of
the primary network on the performance of the secondary network can be relaxed when
the distance between two networks is high. In Figs. 6 and 7, we verify our derivations by
placing the primary user far the secondary network. We can observe from these fig-
ures that the simulation (Ex) and theoretical (As) results are in a good agreement.
Moreover, it can be observed from Fig. 7 that the secrecy performance significantly
enhances when the distance between the relays and the eavesdropper increases (yEincreases).
0 5 10 15 20 2510
−6
10−5
10−4
10−3
10−2
10−1
100
101
γP [dB]
Out
age
Pro
babi
lity
Ex [M=1]An [M=1]As [M=1]Ex [M=2]An [M=2]As [M=2]Ex [M=3]An [M=3]As [M=3]
Fig. 2 The outage probability as a function of �cP in dB when xRi¼ 0:4, xP ¼ yP ¼ 0:4, cth ¼ 1 and K ¼ 1
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
0 5 10 1510−4
10−3
10−2
10−1
100
γP [dB]
Out
age
Pro
babi
lity
Ex [AF−PRS]Ex [DF−PRS]Ex [DF−ORS]Ex [Proposed Protocol]An
Fig. 3 The outage probability as a function of �cP in dB when xRi¼ 0:5, xP ¼ 0, yP ¼ 0:5, cth ¼ 1:5, M ¼ 4
and K ¼ 2
0 2 4 6 8 10 12 14 16 18 200
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
γP [dB]
Erg
odic
Cap
acity
Ex [PU=[0.2 0.2]]An [PU=[0.2 0.2]]Ex [PU=[0.4 0.4]]An [PU=[0.4 0.4]]Ex [PU=[0.6 0.6]]An [PU=[0.6 0.6]]Ex [PU=[0.8 0.8]]An [PU=[0.8 0.8]]
Fig. 4 The average capacity as a function of �cP in dB when xRi2 0:4; 0:6f g, M ¼ 2 and K ¼ 1
D.-B. Ha et al.
123
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
0.5
1
1.5
2
2.5
xP
(yP
)
Sec
recy
Cap
acity Ex [K=1]
Ex [K=2]Ex [K=4]
Fig. 5 The secrecy capacity as a function of xPðyPÞ when xRi2 0:2; 0:4; 0:6; 0:8f g, M ¼ 4, cth ¼ 0:75,
yE ¼ 1, xE ¼ 0:5 and cP ¼ 5dB
0 2 4 6 8 10 12 14 16 18 200
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
γP [dB]
Sec
recy
Cap
acity Ex [K=1]
Ex [K=2]Ex [K=3]As
Fig. 6 The secrecy capacity as a function of �cP in dB when xRi2 0:3; 0:5; 0:7f g, M ¼ 3, cth ¼ 0:75,
yE ¼ 1, xE ¼ 0:5 and xP ¼ yP ¼ 20
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
5 Conclusions
In this paper, the system performance and the secrecy performance of the cooperative relay
networks with K-th best reactive DF relay selection strategy were evaluated. In particular,
we derived the exact expressions for outage probability and ergodic capacity in homo-
geneous and non-homogeneous networks. Furthermore, with the presence of the eaves-
dropper in the considered network, the secrecy performance of the cooperative hop has
been evaluated, in terms of the average secrecy capacity. Finally, computer simulations
have been provided to validate our analysis. The results presented that the performances of
CR significantly depend on the number of relays, the K-best relay selection as well as the
position of the primary user and eavesdropper.
Acknowledgments This research is funded by Vietnam National Foundation for Science and TechnologyDevelopment (NAFOSTED) under Grant Number 102.04-2013.13.
Appendix: A Detailed Derivation of (31)
Using Eqs. (25), (26), and (30), the corresponding asymptotic outage probability of (17) is
given by
~Pout1 �
qp!0 X
Q1;Q2
N\K
YM
t¼K
ksjt ðqPÞM�Kþ1
� 1� expð�ksplÞ þCðM � K þ 2; ksplÞ
ðksplÞM�Kþ1
!
:
ð56Þ
Using again Eqs. (56) and (30) for (19), we get (57) as
0 2 4 6 8 10 12 14 16 18 200.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
γP [dB]
Sec
recy
Cap
acity
Ex [yE
=0.2]Ex [y
E=0.5]
Ex [yE
=1]As
Fig. 7 The secrecy capacity as a function of �cP in dB when xR ¼ 0:5, M ¼ 4, K ¼ 2, cth ¼ 0:5, xE ¼ 1 andxP ¼ yP ¼ 15
D.-B. Ha et al.
123
~Pout2 �
qp!0 X
Q1;Q2
N �K
YM
t¼Nþ1
ksjt 1� exp �kspl� �
þC M � N þ 1; kspl� �
kspl� �M�N
!
�X
c¼1
N
jt¼b
1
N � K þ 1kbd þ
kbd exp �kbpl� �
kbpl
� �
�X
W1;W2
YN
v¼Kþ1
kzvd þkzvd expð�kzvplÞ
kzvpl
� �qM�Kþ1P :
ð57Þ
Thus, combining (56) and (57), we can obtain ~Pout. Similarly, an asymptotic expression for~PH;out in the homogeneous networks can be readily computed as ~PH;out in (31).
References
1. Ghasemi, A., & Sousa, E. S. (2007). Fundamental limits of spectrum-sharing in fading environments.IEEE Transactions on Wireless Communications, 6, 649–658.
2. Duong, T. Q., Yeoh, P. L., Bao, Vo Nguyen Quoc, Elkashlan, M., & Yang, N. (2012). Cognitive relaynetworks with multiple primary transceivers under spectrum sharing. IEEE Signal Processing Letters,19, 741–744.
3. Duy, T. T., & Kong, H. Y. (2014). Adaptive cooperative decode-and-forward transmission with powerallocation under interference constraint. Wireless Personal Communications, 74, 401–414.
4. Bletsas, A., Shin, H., & Win, M. Z. (2007). Cooperative communications with outage-optimal oppor-tunistic relaying. IEEE Transactions on Wireless Communications, 6, 3450–3460.
5. Suraweera, H. A., Smith, P. J., & Shafi, M. (2010). Capacity limits and performance analysis ofcognitive radio with imperfect channel knowledge. IEEE Transactions on Vehicular Technology, 59,1811–1822.
6. Duong, T. Q., da Costa, D. B., Elkashlan, M., & Bao, V. N. Q. (2012). Cognitive amplify-and- forwardrelay networks over Nakagami-m fading. IEEE Transactions on Vehicular Technology, 61, 2368–2374.
7. Kim, K. J., Duong, T. Q., & Tran, X.-N. (2012). Performance analysis of cognitive spectrum-sharingsingle-carrier systems with relay selection. IEEE Transactions on Signal Processing, 60, 6435–6449.
8. Duong, T. Q., Duy, T. T., Matthaiou, M., Tsiftsis, T., & Karagiannidis, G. K. (2013). Cognitivecooperative networks in dual-hop asymmetric fading channels. In Proceedings of IEEE Global Com-munications Conference (Globecom), Atlanta, GA, (pp. 977-983).
9. Bao, V. N. Q., Duong, T. Q., da Costa, D. B., Alexandropoulos, G. C., & Nallanathan, A. (2013).Cognitive amplify-and-forward relaying with best relay selection in non-identical Rayleigh fading.IEEE Communications Letters, 17, 475–478.
10. Sagong, S., Lee, J., & Hong, D. (2011). Capacity of reactive DF scheme in cognitive relay networks.IEEE Transactions on Wireless Communications, 10, 1276–1536.
11. Duy, T. T., & Kong, H. Y. (2013). Performance analysis of incremental amplify-and-forward relayingprotocols with nth best partial relay selection under interference constraint. Wireless Personal Com-munications, 71, 2741–2757.
12. Zhang, X., Yan, Z., Gao, Y., & Wang, W. (2013). On the study of outage performance for cognitiverelay networks (CRN) with the Nth best-relay selection in Rayleigh-fading channels. IEEE Transactionson Wireless Communications, 2, 110–113.
13. Wyner, A. D. (1975). The wire-tap channel. Bell System Techical Journal, 54, 13551367.14. Bloch, M., Barros, J., Rodrigues, M. R. D., & McLaughlin, S. W. (2008). Wireless information theoretic
security. IEEE Transactions on Information Theory, 54, 2515–2534.15. Gopala, P. K., Lai, L., & Gamal, H. E. (2008). On the secrecy capacity of fading channels. IEEE
Transactions on Information Theory, 54, 4687–4698.16. Wang, L., Yang, N., Elkashlan, M., Yeoh, P. L., & Yuan, J. (2014). Physical layer security of maximal
ratio combining in two-wave with diffuse power fading channels. IEEE Transactions on InformationForensics and Security, 9, 247–258.
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123
17. Elkashlan, M., Wang, L., Duong, T. Q., Karagiannidis, G. K., & Nallanathan, A. (2014). On the securityof cognitive radio networks. IEEE Transactions on Vehicular Technology. doi:10.1109/TVT.2014.2358624.
18. Yang, N., Yeoh, P. L., Elkashlan, M., Schober, R., & Collings, I. B. (2013). Transmit antenna selectionfor security enhancement in mimo wiretap channels. IEEE Transactions on Communications, 61,144–154.
19. Yang, N., Yeoh, P. L., Elkashlan, M., Schober, R., & Yuan, J. (2013). MIMO wiretap channels: Securetransmission using transmit antenna selection and receive generalized selection combining. IEEECommunications Letters, 17, 1754–1757.
20. Wang, L., Elkashlan, M., Huang, J., Schober, R., & Mallik, R. K. (2014). Secure transmission withantenna selection in MIMO Nakagami-m channels. IEEE Transactions on Wireless Communications,13, 6054–6067.
21. Wang, L., Elkashlan, M., Huang, J., Tran, N. H., & Duong, T. Q. (2014). Secure transmission withoptimal power allocation in untrusted relay networks. IEEE Wireless Communications Letters, 3,289–292.
22. Wang, L., Kim, K. J., Duong, T. Q., Elkashlan, M., & Poor, H. V. (2015). Security enhancement ofcooperative single carrier systems. IEEE Transactions on Information Forensics and Security, 10,90–103.
23. Liu, Y., Wang, L., Duy, T. T., Elkashlan, M., & Duong, T. Q. (2015). Relay selection for securityenhancement in cognitive relay networks. IEEE Wireless Communications Letters, 4, 46–49.
24. Duy, T. T., & Kong, H. Y. (2013). Performance analysis of two-way hybrid decode-and-amplifyrelaying scheme with relay selection for secondary spectrum access. Wireless Personal Communica-tions, 69, 857–878.
25. Lee, J., Wang, H., Andrews, J. G., & Hong, D. (2011). Outage probability of cognitive relay networkswith interference constraints. IEEE Transactions on Wireless Communications, 10, 390395.
26. Gradshteyn, I., & Ryzhik, I. (2007). Table of integrals, series, and products (7th ed.). New York,London: Academic Press Inc.
27. Duong, T. Q., Duy, T. T., Elkashlan, M., Tran, Nghi. H., & Dobre, Octavia A. (2014). Securedcooperative cognitive radio networks with relay selection. In Proceedings of IEEE Globecom2014,Austin, TX USA, (pp. 3151-3156).
28. Bao, V. N. Q., & Duy, T. T. (2012). Performance analysis of cognitive underlay DF relay protocol withKth best partial relay selection. In Proceedings of ATC2012, Ha Noi, VietNam, (pp. 130-135).
29. Duy, T. T., Duong, T. Q., da Costa, D. B., Bao, V. N. Q., & Elkashlan, M. (2015). Proactive relayselection with joint impact of hardware impairment and co-channel interference. IEEE Transactions onCommunications, 63, 1594–1606.
Dac-Binh Ha received the B.S. degree in Radio Technique, the M.Sc.and Ph.D. degree in Communication and Information System fromHuazhong University of Science and Technology (HUST), China in1997, 2006, and 2009, respectively. He is currently the dean of Facultyof Electrical and Electronics Engineering, Duy Tan University, DaNang, Vietnam. During 1997–2001, he worked for Center of TelecomTechnology, Vietnam. From 2001 to 2003, he joined the Faculty ofElectrical and Electronics Telecommunications Engineering of Ho ChiMinh City University of Transport, Vietnam. He is the recipient of theChinese Government Distinguished International Students Scholarship.His research interests are secrecy physical layer communications,MIMO systems, combining techniques, cooperative communications,cognitive radio, and image processing. Dr. Ha is a member of theInstitute of Electrical and Electronics Engineers (IEEE). He is arecipient of the Best Paper Award at the 2014 IEEE InternationalConference on Computing, Management and Telecommunications.
D.-B. Ha et al.
123
Tung Thanh Vu received the B.Sc. degree in Telecommunicationsand Networking from the Ho Chi Minh City University of Science in2012. He is pursuing his M.Sc. degree in Telecommunications Engi-neering at the Ho Chi Minh City University of Technology. His currentresearch interests include cognitive ratio networks, cooperative net-works, physical layer security, interference alignment and optimizationdesign.
Tran Trung Duy was born in Nha Trang city, Vietnam, in 1984. Hereceived the B.E. degree in Electronics and TelecommunicationsEngineering from the French-Vietnamese training program for excel-lent engineers (PFIEV), Ho Chi Minh City University of Technology,Vietnam in 2007. In 2013, he received the Ph.D. degree in electricalengineering from University of Ulsan, South Korea. In 2013, he joinedthe Department of Telecommunications, Posts and Telecommunica-tions Institute of Technology (PTIT), as a lecturer. His major researchinterests are cooperative communications, cognitive radio, and physi-cal layer security.
Vo Nguyen Quoc Bao was born in Khanh Hoa, Vietnam, in 1979. Hereceived the B.E. and M.Eng. degree in electrical engineering from HoChi Minh City University of Technology (HCMUT), Vietnam, in 2002and 2005, respectively, and Ph.D. degree in electrical engineering fromUniversity of Ulsan, South Korea, in 2010. In 2002, he joined theDepartment of Electrical Engineering, Posts and TelecommunicationsInstitute of Technology (PTIT), as a lecturer. Since February 2010, hehas been with the Department of Telecommunications, PTIT, where heis currently an Assistant Professor. His major research interests aremodulation and coding techniques, MIMO system, combining tech-nique, cooperative communications and cognitive radio. Dr. Bao is aMember of Korea Information and Communications Society (KICS),The Institute of Electronics, Information and Communication Engi-neers (IEICE) and the Institute of Electrical and Electronics Engineers(IEEE).
Secure Cognitive Reactive Decode-and-Forward Relay Networks...
123