An Analytical Model for Primary User Emulation
Attacks in Cognitive Radio Networks
S. Anand, Z. Jin and K. P. Subbalakshmi
Department of Electrical and Computer Engineering
Stevens Institute of Technology, New Jersey, USA
Abstract—In this paper, we study the denial-of-service (DoS)
attack on secondary users in a cognitive radio network by
primary user emulation (PUE). Most approaches in the literature
on primary user emulation attacks (PUEA) discuss mechanisms
to deal with the attacks but not analytical models. Simulation
studies and results from test beds have been presented but
no analytical model relating the various parameters that could
cause a PUE attack has been proposed and studied. We propose
an analytical approach based on Fenton’s approximation and
Markov inequality and obtain a lower bound on the probability
of a successful PUEA on a secondary user by a set of co-operating
malicious users. We consider a fading wireless environment and
discuss the various parameters that can affect the feasibility of
a PUEA. We show that the probability of a successful PUEA
increases with the distance between the primary transmitter and
secondary users. This is the first analytical treatment to study
the feasibility of a PUEA.
Keywords – Cognitive radio networks malicious user, primary user
emulation attack
I. INTRODUCTION
Spectrum sharing has always been an important aspect of
system design in wireless communication systems due to the
scarcity of the available resources/spectrum. Cognitive radio
networks [1] enable usage of unused spectrum in a network,
A, by users belonging to another network, B. These users
thereby become “secondary users” to the network A. The
users that originally subscribed to the network A are called
“primary users” of network A. One example of cognitive radio
network is the usage of white spaces (or unused spectrum) in
the television (TV) band. The TV transmitter then becomes
a primary transmitter and TV receivers are primary receivers.
Other users who are not TV subscribers but wish to use the
white spaces in the TV band for their own communication
become secondary transmitters/receivers. The IEEE 802.22
working group on wireless regional area networks (WRAN)
[2] provide the physical layer (PHY) and medium access
control (MAC) specifications for usage of the TV white
spaces. More details on the IEEE 802.22 can be found in [3],
[4]. The developments in software defined radio (SDR) [5]
enables implementation of re-configurable MAC for dynamic
spectrum access (DSA). Akyildiz et al [6] provide a detailed
survey of the developments in SDR, DSA and cognitive
radio. The etiquette followed in cognitive radios is that the
secondary users evacuate the used spectrum once they detect
This work was funded by a research grant from NSF Cyber Trust Grant
No. 0627688
a primary transmission. In [6], the authors also provide a
detailed description of the different sensing mechanisms that
enable secondary users to detect the presence of a primary user
namely: (a) Transmitter detection, (b)co-operative detection
and (c) interference-based detection. Transmitter detection, in
turn, can be performed using one of three mechanisms namely:
(i) matched filter detection, (ii) energy detection and (iii)
cyclostationary feature detection. A detailed description and
comparative study of the above methods are also provided in
[6]. Protocols for sensing primary transmission and evacuating
the spectrum were discussed by Visotsky et al [4] and by Liu
and Ding [7].
The etiquette of spectrum evacuation could however result
in denial-of-service attacks on secondary users if the system
is not carefully designed. This is explained as follows. Con-
sider a set of secondary users in the system. A subset of
users could forge the essential signal characteristics of the
primary and generate enough power at the good secondary
user locations to confuse the secondaries into thinking that a
primary transmission is under way. The secondaries obeying
the normal etiquette will vacate the spectrum unnecessarily.
The subset of users would then use the evacuated white space
for themselves. The secondary users who transmit to emulate
the primary transmitter are referred to as “malicious users”
while the other secondary users who evacuate the spectrum
upon sensing the transmission from the primary transmitter
or the malicious users are termed as “good” secondary users
1
.
Such an attack by malicious users on secondary users is called
a primary user emulation attack (PUEA). It is noted that such
attacks could lead to big disadvantages because several good
users could lose access to the network due to the presence of
a few malicious users. This, in turn, leads to poor usage of
spectrum for authorized users and an unfair advantage for the
malicious users.
PUEA in cognitive radio networks was studied in
[8],[9],[10]. In [8], Chen and Park propose two mechanisms
to detect a PUEA namely the distance ratio test (DRT) and
the distance difference test (DDT), which use the ratio and
the difference, respectively, of the distances of the primary
and malicious transmitters from the secondary user to detect
a PUEA. In [9], Chen et al discuss defense against PUEA
by localization of primary transmitters. Directional antennas
were proposed to determine the angle of arrival of the primary
1
Henceforth, throughout the paper, whenever we mention “secondary
users”, we refer to “good secondary users” unless explicitly mentioned
otherwise.
signal, and using this, the time of arrival and the received
signal strength, the secondary users determine the location
of the primary transmitter. A different kind of threat albeit
not directly a PUEA, was discussed by Chen et al in [10].
The authors consider a system where spectrum sensing is
done and a hypothesis testing method is used to detect a
transmission, which in the case of cognitive radio networks
could be a primary transmission. A Byzantine failure model
due to fraudulent reporting of spectrum sensing was discussed
and a weighted sequential ratio test was proposed to overcome
this attack.
In most approaches, the detection of PUEA depends on
the determination of the location of the primary transmitter,
which, in turn, depends on the direction of signal arrival.
The dependence on the directionality of the antennas at the
receiver makes the detection process complex because most
of the incumbent receivers in wireless and cellular networks
use omni directional antennas.
We present the first ever analytical treatment of the feasi-
bility of a PUEA. We derive mathematical expressions for
the probability of a successful PUEA and provide lower
bounds on the probability of a successful attack using Fenton’s
approximation and Markov inequality. We consider a wire-
less environment with losses due to attenuation, fading and
shadowing. We consider a variation of the energy detection
mechanism mentioned in [6]. We model the received power at
a secondary user as a log-normally distributed random variable
and use Fenton’s approximation to determine the mean and the
variance of the received power. We then use the value of the
derived mean and variance to determine a lower bound on the
probability of a successful PUEA using Markov inequality. We
discuss the various parameters that can affect the feasibility of
a PUEA. We show that the probability of a successful PUEA
increases with the distance between the primary transmitter
and secondary users. The rest of the paper is organized as
follows. In Section II, we present the system model. Section III
presents the analytical model for the probability of a successful
PUEA. In Section IV, we present the numerical results and
discussion. Section V presents the conclusion.
II. SYSTEM MODEL
Consider a system as shown in Fig. 1. All secondary and
malicious users are distributed in a circular grid of radius
R. A primary transmitter is present at a distance of at least
D
p
from all the users. The energy detection method for
spectrum sensing by secondary users in as follows [6]. Each
secondary user measures the energy of the received signal and
compares the measured energy with a pre-set threshold, Λ. If
the measured energy is greater than Λ, then the secondary
user concludes that a primary transmission is present. Else,
the secondary user concludes that the spectrum is free for
usage. We consider a variation of this method for spectrum
sensing, where each secondary user measures the received
power and compares them with two thresholds, ǫ
l
and ǫ
h
.
If the measured signal power lies between ǫ
l
and ǫ
h
, then
the secondary user concludes that a primary transmission is
present and refrains from using the spectrum. Otherwise, the
secondary users concludes that there exists a white space.
The reason for such a mechanism is that the measurement
threshold for typical cognitive radio system is -93 dBm [2].
If the measurement is based on a single energy threshold,
then even a single malicious user transmitting at sufficiently
large power can cause a successful PUEA. In this case also, a
set of malicious users can transmit in such a way that the
total received power at a good secondary user due to the
transmission by all the malicious users is very close to that due
to the transmission from the primary transmitter, thus resulting
in a primary user emulation attack (PUEA). A successful
PUEA is defined as the event that the absolute difference
between the received powers from the primary and that from
all the malicious users is below a specified threshold, ǫ. It is
of interest to determine the probability of a successful PUEA
at any secondary user. We make the following assumptions for
our analysis.
• There are M malicious users and N good secondary users
in the system.
• The primary transmitter is at a minimum distance of D
p
from all the users.
• The primary transmitter transmits at a power P
t
.
• The malicious users transmit at a power P
m
. (Typically,
P
m
<< P
t
).
• The positions of the good and malicious users are uni-
formly distributed in the circular grid of radius R.
• The co-ordinates
2
of the primary transmitter are fixed at
a point (r
p
, θ
p
) and this position is known to all the users
in the grid.
• The positions of the good users and the malicious users
are statistically independent of each other.
• The RF signals from the primary transmitter and the
malicious users undergo path loss, log-normal shadowing
and Rayleigh fading.
• The shadowing random variable from the primary trans-
mitter to the i
th
secondary user is
G
(i)
p
2
= 10
ξ
(i)
p
10
,
where ξ
(i)
p
∼ N
0, σ
2
p
.
• The shadowing random variable from the j
th
malicious
user to the i
th
secondary user is (G
ij
)
2
= 10
ξ
ij
10
, where
ξ
ij
∼ N
0, σ
2
m
.
• The Rayleigh fading random variables from the primary
transmitter and all malicious users to all secondary users
are identically distributed with mean ∆.
• We consider a free space propagation model for the signal
from the primary transmitter and a two-ray ground model
for the signal from the malicious users thus resulting in
a path loss exponent of 2 for the propagation from the
primary transmitter and a path loss exponent of 4 for the
propagation from the malicious users. This is because, the
primary transmitter is so far away from the secondary and
malicious users that the signal due to multi-path can be
neglected. However, the distances from malicious users
are not large enough to ignore the effects of multi-path.
• For any secondary user fixed at co-ordinates (r, θ), no
2
Throughout this paper, whenever we mention “co-ordinates” we mean
“polar co-ordinates” unless explicitly mentioned otherwise.
malicious users are present within a circle of radius R
0
centered ar (rθ). If this restriction is not posted, then
the power received due to transmission from any subset
of malicious users present within this grid will be much
larger than that due to a transmission from a primary
transmitter thus resulting in a failed PUEA all the time.
On the other hand, if the malicious users deploy power
control, then the malicious user present in this grid can
modify its transmit power in such a way so that the PUEA
is successful all the time. The distance R
0
is called the
“exclusive distance from the secondary user”.
R
0
Good Secondary Users
Malicious Users
p
D
R
Primary Transmitter
Fig. 1. A typical cognitive radio network in a circular grid with secondary
and malicious users.
III. ANALYTICAL MODEL
The received power at the i
th
secondary user from the
primary transmitter, P
(p)
r
(i), is given by
P
(p)
r
(i) = P
t
d
(i)
p
−2
G
(i)
p
2
R
(i)
p
2
, (1)
where d
(i)
p
is the distance from the primary transmitter to the
i
th
secondary user,
G
(i)
p
2
is the log-normal shadowing from
the primary transmitter to the i
th
secondary user and R
(i)
p
is
the Rayleigh fading from the primary transmitter to the i
th
secondary user. The received power at the i
th
secondary user
due to the transmission from all the malicious users, P
(m)
r
(i),
is given by
P
(m)
r
(i) =
M
X
j=1
P
m
d
−4
ij
(G
ij
)
2
(R
ij
)
2
, (2)
where d
ij
is the distance from the j
th
malicious user to the
i
th
secondary user, (G
ij
)
2
is the log-normal shadowing from
the j
th
malicious user to the i
th
secondary user and R
ij
is
the Rayleigh fading from the j
th
malicious user to the i
th
secondary user. A PUEA on the i
th
secondary user is deemed
successful if for a specified threshold, ǫ,
P
(p)
r
(i) − P
(m)
r
(i)
< ǫ. (3)
The probability of a successful PUEA on the i
th
secondary
user is given by
p
P UEA
= Pr
n
P
(p)
r
(i) − P
(m)
r
(i)
< ǫ
o
. (4)
Conditioned on the positions of the secondary and malicious
users and the Rayleigh fading terms from the primary and all
the malicious users, P
(p)
r
(i) and each term in the summation of
the right hand side in Eqn. (2) are log-normally distributed ran-
dom variables. P
(m)
r
(i) can be approximated as a log-normally
distributed random variable whose mean and variance can be
obtained by using Fenton’s method [11]. A detailed description
of Fenton’s method is provided in Appendix I.
Let P
diff
r
(i)
△
= P
(p)
r
(i) − P
(m)
r
(i). The random variable
P
diff
r
(i) is modeled as a log-normally distributed random
variable of the form P
diff
r
(i) = 10
ω
d
(i)
10
, where ω
d
(i) ∼
N
µ
d
(i), σ
2
d
(i)
. Fenton’s method needs to be applied again
to obtain the values of µ
d
(i) and σ
2
d
(i). Conditioned on the
Rayleigh fading random variables from the primary and all the
malicious users to the secondary user i and the positions of
the secondary user and all the malicious users, the probability
of a successful PUEA, ˆp
P UEA
, can be obtained as
ˆp
P UEA
= 1 − Q
ǫ
dB
+ µ
d
(i)
σ
d
(i)
− Q
ǫ
dB
− µ
d
(i)
σ
d
(i)
, (5)
where ǫ
dB
is the threshold ǫ expressed in decibels (i. e., ǫ
dB
=
10 log
10
ǫ) and Q(x) =
1
√
2π
R
∞
x
e
−
y
2
2
dy. The probability
of a successful PUEA, p
P UEA
defined in Eqn. (4) can be
obtained by averaging ˆp
P UEA
in Eqn. (5) over the positions
of the secondary and malicious users and the Rayleigh fading
from the primary and all the malicious users to the secondary
user. For M malicious users, this results in 2M integrations
corresponding to the positions of the malicious users (since
each position has two co-ordinates), two integrations corre-
sponding to the position of the secondary user, M integrations
corresponding to the Rayleigh fading from all the malicious
users to the secondary user and one integration corresponding
to the Rayleigh fading from the primary transmitter to the
secondary user. Thus, a total of 3(M + 1) integrations needs
to be performed for M malicious users. Therefore, exact
evaluation of the probability in Eqn. (4) is very complex.
Hence, we use the Markov inequality [12] to bound the
probability.
Consider a random variable X such that Pr{X < 0} = 0.
For any α > 0, the Markov inequality is [12]
Pr{X > α} ≤
E[X]
α
. (6)
Using this, the probability p
P UEA
in Eqn. (4) can be bounded
as
p
P UEA
≥ 1 −
E
h
P
(p)
r
(i)
i
− E
h
P
(m)
r
(i)
i
ǫ
. (7)
To evaluate the expectations in the above, we adopt an
approach based on Fenton’s approximation. This is described
in detail as follows. As mentioned earlier, conditioned on
the position of the secondary user, the received power at
a secondary user due to the primary transmission is a log-
normally distributed random variable. The mean E
h
P
(p)
r
(i)
i
is then given by
E
h
P
(p)
r
(i)
i
= P
t
∆e
1
2
a
2
σ
2
p
E
d
(i)
p
−2
, (8)
where a =
ln 10
10
. The distance d
(i)
p
is given by
d
(i)
p
=
q
r
2
i
+ r
2
p
− 2r
i
r
p
cos(θ
i
− θ
p
), (9)
where (r
i
, θ
i
) are the co-ordinates of the i
th
secondary user
and (r
p
, θ
p
) are the co-ordinates of the primary transmitter.
The expectation, E
d
(i)
p
−2
, in Eqn. (8) can be evaluated
as
E
d
(i)
p
−2
=
1
πR
2
Z
R
r
i
=0
Z
2π
θ
i
=0
r
i
dr
i
dθ
i
r
2
i
+ r
2
p
− 2r
i
r
p
cos(θ
i
− θ
p
)
. (10)
The expression in Eqn. (10) is substituted in Eqn. (8) to obtain
the value of E
h
P
(p)
r
(i)
i
.
To evaluate E
h
P
(m)
r
(i)
i
we first note that conditioned on
the locations of the malicious and secondary users and the
Rayleigh fading term, each term in the summation of Eqn.
(2) is a log-normally distributed random variable of the form
10
ξ
ij
10
= e
aξ
ij
, where ξ
ij
∼N (µ
ij
, σ
2
m
), where µ
ij
is given
by
µ
ij
= P
dB
m
+ 10 log
10
∆ − 20 log
10
d
2
ij
, (11)
where P
dB
m
is the transmit power from the malicious users
represented in decibels (i. e., 10 log
10
P
m
) and d
ij
is given by
Eqn. (9) by replacing r
p
and θ
p
by r
j
and θ
j
, repespectively.
We then approximate the sum of the log-normally distributed
random variables (in the right hand side of Eqn. (2)) to
be a log-normally distributed random variable of the form
10
ω
M
10
= e
aω
M
, where ω
M
i
∼ N (ˆµ, ˆσ
2
), by using Fenton’s
approximation. Conditioned on the locations of the malicious
and the secondary users, ˆσ
2
and ˆµ can be obtained as
3
ˆσ
2
=
1
a
2
ln
1 +
e
a
2
σ
2
m
− 1
P
M
j=1
e
2aµ
ij
P
M
j=1
e
aµ
ij
2
, (12)
and
ˆµ =
1
a
ln
M
X
j=1
e
aµ
ij
−
a
2
ˆσ
2
− σ
2
m
. (13)
It is essential to average over the positions of the malicious
and secondary users to obtain the mean and variance of ω
M
.
This would involve integrating the expressions in Eqn. (12)
and Eqn. (13) over r
j
, θ
j
for j = 1, 2, 3, . . . , M and r
i
and
θ
i
, thus resulting in 2(M + 1) integrations. Although this is
smaller than the number of integrations required to obtain the
exact value or p
P UEA
, it still remains too complex to evaluate.
In order to reduce the complexity of the computations, we
make two modifications to the analysis:
1) Without loss of generality, we fix the position of the
secondary user at (0, 0)
4
.
3
The detailed derivation for the expressions in Eqns. (12)- (15) can be
obtained by following the description provided in Appendix I.
4
For any other fixed position of the secondary user, the analysis would still
be valid by making a suitable co-ordinate transformation.
2) We approximate the received power at a secondary user
from each of the malicious users to be independent
and identically distributed. This is valid due to the
symmetry of the system and the fact that the malicious
users can be present uniformly in an annular region
between the circles centered at (0, 0) and radii R
0
and
R. Such approximations for analysis of other parameters
in cognitive radio networks were made in [13],[14],[15].
Using the above modifications, ˆσ
2
and ˆµ can be obtained as
ˆσ
2
=
1
a
2
ln
1 +
(e
a
2
σ
2
m
− 1
M
, (14)
and
ˆµ = µ
ij
+
1
a
ln M −
a
2
ˆσ
2
− σ
2
m
. (15)
The expectation E
h
P
(m)
r
(i)
i
can then be obtained as
E
h
P
(m)
r
(i)
i
=
1
π(R
2
− R
2
0
)
e
1
2
a
2
ˆσ
2
Z
2π
θ
j
=0
Z
R
r
j
=R
0
e
aˆµ
r
j
dr
j
dθ
j
. (16)
Using the expression for distance between two points in polar
co-ordinates given by Eqn. (9), the above can be simplified
and obtained as
E
h
P
(m)
r
(i)
i
= M
1
2R
2
0
(R
2
− R
2
0
)
e
1
2
a
2
ˆσ
2
∆. (17)
By fixing the co-ordinates of a secondary user at (0, 0),
E
h
P
(p)
r
(i)
i
is obtained by removing the integration in Eqn.
(8) as
E
h
P
(p)
r
(i)
i
=
P
t
e
1
2
a
2
σ
2
p
∆
r
2
p
. (18)
The expression for E
h
P
(m)
r
(i)
i
from Eqn. (17) and
E
h
P
(p)
r
(i)
i
from Eqn. (18) are substituted in Eqn. (7) to
obtain the lower bound on a succesful PUEA on a secondary
user.
IV. RESULTS AND DISCUSSION
We consider the following values of the system parameters
for our numerical computations. We consider σ
p
= 8 and
σ
m
= 5.5, by assuming urban and suburban environments
for the propagations from the primary transmitter and mali-
cious users, respectively [16]. We consider the mean Rayleigh
fading, ∆, to be unity. The transmit power from the malicious
users, P
m
, is taken to be 4 Watts as in [9].
Fig. 2 presents the lower bound on the probability of a
succesful PUEA obtained by the analysis in Section III, in a
system with 100 malicious users when the primary transmitter
is at a distance of 2000m from the secondary user. The
threshold values of 0.1, 0.05 and 0.025 shown in Fig. 2
correspond to a difference of 100mW, 50mW and 25mW,
respectively, between the received powers from the primary
transmitter and that from the malicious users. The thresholds
are chosen based on the following argument. For a primary
transmitter 2000m away from the secondary user, the received
power at the receivers vary typically between 0.1mW to 7.5W
with mean 150mW (this is from the fact that a Gaussian
random variable X ∼ N (µ, σ
2
) typically takes values between
µ − 3σ and µ + 3σ). Hence, a difference of 100mW or lesser
can be considered a succesful PUEA.
It is noted that the plots only present a lower bound and the
actual probability may be higher than that shown. It is noted
that for small values of R
0
, the lower bound is 0. This is
because, for smaller values of R
0
the malicious users are too
close to the secondary user and when transmitting at maximum
power of 4 Watts each, they result in a very large received
power at the secondary user, thereby making the secondary
user able to differentiate between a primary transmission and
a malicious transmission. As expected, when the threshold
reduces, the lower bound becomes looser (i.e., the lower bound
decreases).
Fig. 3 shows the lower bound when the primary transmitter
is at a distance of 8000m from the secondary user. In this
case, it is observed that for sufficiently large R
0
(i. e., R
0
>
90m), even a threshold of 0.01 (i.e., 10mW) of the differences
between the received powers due to primary and malicious
transmissions results in a significant probability of a succesful
PUEA.
The following inferences can be made from Figs. 2 and 3.
1) Since small values of R
0
result in a large received power
at the secondary user due to transmission from malicious
users, very large values of R
0
may also result in low
PUEA since the received powers at the secondary users
due to transmission from malicious users may be too
small. One can then find a range of R
0
in which an
attack can be succesful.
2) The significantly high values of a succesful PUEA under
the absence of any power control at the malicious users
indicate that with suitable power control, the probability
of a succesful PUEA can further be enhanced. In par-
ticular, it is possible to obtain a set of transmit powers
for each of the malicious users such that the probability
of a succesful PUEA at a secondary user is 1.
V. CONCLUSION
We proposed an analytical approach and obtained a lower
bound on the probability of a successful PUEA on a secondary
user in a cognitive radio network by a set of co-operating mali-
cious users. We show that the probability of a succesful PUEA
increases with the distance between the primary transmitter
and secondary users. This is the first analytical treatment to
study the feasibility of a PUEA. We showed that our bounds
enable in obtaining insights on possible ranges of exclusive
regions in which an attack is most likely. Our results motivate
the study of energy efficient PUEA attacks. Extension of our
approach to determine the lower bounds for the probability of
successful PUEA in systems deploying other spectrum sensing
mechanisms described in [6] is a topic for further investigation.
30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Exclusive Distance From Secondary (R
0
)
Lower Bound on Probability of Primary Emulation Attack
Threshold=0.1
Threshold=0.05
Threshold=0.025
Fig. 2. Lower bound on the probability of a succesful PUEA when the
primary transmitter is at a distance of 2 Km from the secondary user.
30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Exclusive Distance from Secondary (R
0
)
Lower Bound on Probability of Primary Emulation Attack
Threshold=0.1
Threshold=0.05
Threshold=0.025
Threshold=0.01
Fig. 3. Lower bound on the probability of a succesful PUEA when the
primary transmitter is at a distance of 8 Km from the secondary user.
REFERENCES
[1] S. Haykin, “Cognitive radio: Brain empowered wireless communica-
tions,” IEEE Jl. on Sel. Areas in Commun., vol. 23, no. 2, pp. 201–220,
Feb. 2005.
[2] “IEEE Standards for information technology- Telecommunications
and information exchange between systems- Wireless Regional Area
Networks-Specific Requirements- Part 22-Cognitive wireless RAN
medium access control (MAC) and physical layer (PHY) specifications:
Policies and procedures for operation in the TV bands,” Jun. 2006.
[3] C. Cordeiro, K. Challapali, D. Birru, and S. Shankar, “Ieee 802.22: The
first worldwide wireless standard based on cognitive radios,” Proc., IEEE
Symposium of New Frontiers in Dynamic Spectrum Access Networks
(DySPAN) 2005, pp. 328–337, Nov. 2005.
[4] E. Visotsky, S. Kuffner, and R. Peterson, “On collaborative detection of
tv transmission in support of dynamic spectrum sharing,” pp. 338–345,
Nov. 2005.
[5] A. Harrington, C. Hong, and T. Piazza, “Software defined radio: The
revolution of wireless communication,” White paper, Ball State Univer-
sity. [Online]. Available: http://www.bsu.edu/cics/alumni/whitepapers/
[6] I. F. Akyildiz, W. Lee, M. C. Vuran, and S. Mohanty, “Next gen-
eration/dynamic spectrum access/cognitive radio: A survey,” Elsevier
Journal on Computer Networks, vol. 50, pp. 2127–2158, May 2006.
[7] X. Liu and Z. Ding, “ESCAPE: a channel evacuation protocol for
spectrum-agile networks,” Proc., IEEE Symposium of New Frontiers in
Dynamic Spectrum Access Networks (DySPAN) 2007, pp. 292–302, Apr.
2007.
[8] R. Chen and J. M. Park, “Ensuring trustworthy spectrum sensing
