A Two Step Secure Spectrum Sensing Algorithm Using Fuzzy Logic for Cognitive Radio Networks

doi:10.4236/ijcns.2011.48062

Paper Menu >>

Journal Menu >>

Int. J. Communications, Network and System Sciences, 2011, 4, 507-513

doi:10.4236/ijcns.2011.48062 Published Online August 2011 (http://www.SciRP.org/journal/ijcns)

A Two Step Secure Spectrum Sensing Algorithm Using

Fuzzy Logic for Cognitive Radio Networks

Ehsan MoeenTaghavi, Bahman Abolhassani

School of Electrical Engineering, Iran University of Scien c e and Technology (IUST), Tehran, Iran

E-mail: [email protected], [email protected]

Received June 5, 2011; revised July 9, 2011; accepted July 18, 2011

Abstract

In this paper, a two step secure spectrum sensing algorithm is proposed for cognitive radio networks. In this

algorithm, the sensing results of secondary users are pre-filtered and applying fuzzy logic, so, the overall

sensing performance of the network is improved. To determine pre-filter parameters, statistical parameters of

the sensing results are used to remove those sensing results which are far from the majority sensing results.

However, to obtain a better performance in the spectrum sensing, we propose a fuzzy logic to nullify the ef-

fects of malicious users who transmit false sensing data to the fusion center. We further propose a Fuzzy

Trust Level for each user as to weight the sensing result of the corresponding user before combining all

sensing results in the fusion center. Simulation results demonstrate that our proposed algorithm yield signifi-

cant improvement in the performance of the spectrum sensing and identifying malicious users.

Keywords: Cognitive Radio, Cooperative Spectrum Sensing, Energy Detection, Malicious User Detection,

Pre-filtering, Fuzzy Logic

1. Introduction

Traditionally, fixed spectrum bands have been assigned

to specific services for a long time. This policy has led to

inefficient spectrum usage, and caused cognitive radio

networks were proposed. In these networks, cognitive

radios, which are called secondary (unlicensed) users are

allowed to use the primary (licensed) users’ bands when

such bands are unoccupied by primary users (Pus).

However, the secondary users (SUs) must make the band

vacant immediately after a PU starts transmitting in the

corresponding band [1]. Therefore, the most important

task of a cognitive radio is spectrum sensing. Spectrum

sensing techniques include energy detection, cyclosta-

tioary feature detection and matched filter detection [2-4].

The performance of a spectrum sensing is determined by

two probabilities: probability of detection (Pd) and prob-

ability of false alarm (Pf). Pd is the pro bability of declar-

ing the chann el is occupied while the PU is present. Pf is

the probability of declaring the channel is occupied by a

PU while the PU has no transmission. Among different

spectrum sensing schemes for reliably identifying the

licensed spectrum status, the energy detector scheme

incurs a very low implementation cost and therefore is

widely used. It serves as the optimal method to detect the

signal transmitted by a PU whose location is unknown,

and the detector only knows the power of the received

signal [6]. The problem of this scheme is that the re-

ceived signal power can be seriously weakened at a par-

ticular geographical location du e to multipath fading and

shadowing effects [7]. In these circumstances, it is diffi-

cult for a single sensing user to distinguish between an

idle band and a deep faded one. In order to overcome this

problem, cooperative spectrum sensing schemes have

been proposed [5,8,9]. However, in cooperative sensing,

due to imperfect channel between a primary user and a

secondary user (SU) or dishonestly behavior of a SU, a

user might send false sensing result to the fusion center.

So, the performance of the system degrades severely. To

overcome this problem, secure spectrum sensing has

been proposed.

In [8], a spectrum sensing data falsification problem

was solved by weighted sequential probability ratio test

(WSPRT), which gives a good performance. However,

this method requires the knowledge of locations of sens-

ing terminals and position of PU for obtaining some re-

quired prior probabilities. This is inappropriate for mo-

bile cognitive radios and for systems in which the loca-

tion of the primary user is completely unknown. In [10],

a robust secure distributed spectrum sensing scheme is

E. MOEENTAGHAVI ET AL.

508

proposed that uses robust statistics to approximate the

distribution for both hypotheses of all users, discriminat-

ingly, based on their past data report. Th e authors in [11]

propose the majority rule in the fusion center to nullify

the effects of the malicious users. In [5], an effective

weighted combining method is proposed to reduce the

impact of false information. In [12], a defense scheme is

proposed that computes suspicious levels and trust values

of the users. In our previous work [13], malicious user

detection based on outlier energy detection techniques is

proposed and a filtering method is used based on statis-

tical parameters of sensing results to eliminate the effects

of malicious users. In this paper, we propose a two step

secure spectrum sensing algorithm. At first, based on the

statistical parameters of SUs sensing results, a pre-filter

is designed to remove the sensing results of the secon-

dary users which are far from the others. Then, trust

weighted values are assigned to the users whose sensing

results are passed from the filter based on the fuzzy logic.

Finally, a weighted combining method is proposed to

make final decision in the fusion center.

The rest of the paper is organized as follows: In Sec-

tion 2, the system model is described and a brief back-

ground about fuzzy logic is presented. In Section 3, our

new algorithm to nullify the effects of malicious users is

proposed. Simulation results are carried and analyzed in

Section 4, and conclusions are drawn in Section 5.

2. System Description

2.1. System Model

We discuss a cognitive radio network consist of one PU

and a group of N SUs. A cooperative spectrum sensing is

considered in which channels between a primary user

and each secondary user have independent and identi-

cally distributed (i.i.d) Rayleigh distributions. Variation

in path-loss is neglected. Each SU conducts energy de-

tection and transmits the measured value of received

signal energy in a perfect (i.e. error free) control channel

to the fusion center. By combining the sensing results

received from different SUs, the fusion center makes the

final decision regarding the presence or absence of the

primary user. If en[k] for n=1,2, ···, N represents the re-

ceived signal energy of nth SU at time instant k, hy-

potheses H1 and H0 denote the presence and absence of

the primary signal, respectively. Then, the signal energy

received by nth SU is given by:



 



nTT

htstztt H

zt tH















(1)

where T represents the time interv al of sensing, s(t) is the

primary signal, hn(t) denotes the channel gain between

the primary user and the nth secondary user, and zn(t) is

the additive white Gaussian noise (AWGN).

A threat to cooperative spectrum sensing is transmis-

sion of false results by attackers to the fusion center,

which leads to make a wrong decision (SSDF1 attack).

So, cooperative spectrum sensing in adversarial envi-

ronments where a malicious user sends false data de-

grades the performance of the system severely. In adver-

sarial environments, different kinds of malicious users

can affect the spectrum sensing system. They may send

data indicating the presence of the PU to the fusion cen-

ter (“Always Yes” malicious users). These malicious

users cause the fusion center to erroneously decide that

the PU is present. Then, malicious users selfishly use the

entire free spectrum band and also probability of false

alarm is increased. Another kind of malicious are those

who always send data indicating the absence of the pri-

mary user (“Always No” malicious users). This kind of

malicious users causes the interference among the pri-

mary and secondary user’s signal and decrease probabil-

ity of detection [14]. Therefore, the fusion scheme must

be robust enough. In this paper, we propose a two step

algorithm for secure spectrum sensing. In step 1, the

sensing results in the data collector are pre-filtered and in

step 2, we assign each user a Fuzzy Trust Level using

fuzzy logic. In our proposed algorithm, more reliable

users are assigned with higher trust levels. This algo-

rithm also detects suspicious users with their corre-

sponding suspiciou s levels using fuzzy logic. Finally, the

sensing results are combined together based on their

Fuzzy Trust Levels in the fusion center.

2.2. Overview of Fuzzy Logic

In this section, we present a brief background on fuzzy

logic. The reason for that is because our proposed algo-

rithm integrates fuzzy logic with spectrum sensing in

order to better detect malicious user.

Fuzzy logic was introduced by Dr. Lotfi Zadeh of

UC/Berkeley in the 1960’s as a mean to model the un-

certainty of natural language. fuzzy logic, a widely de-

ployed technology for developing sophisticated control

systems [15,16], provides a simple way to get definite

precise conclusion and solution based on unclear, impre-

cise, ambiguous or missing input information. Figure 1

shows the steps that fuzzy logic controller is composed

of. The steps of a fuzzy logic can be summarized as fol-

lows: 1) receiving input values representing measure-

ments of the parameters to be analyzed; 2) subj ecting the

input value to if-then fuzzy rules; 3) averaging and

1Spectrum Sensing D ata Falsification Attack.

E. MOEENTAGHAVI ET AL.

509

Figure 1. Components of the fuzzy logic controller [17].













 

med1.5 ,

med1.5 .

ekk k













 (2)

weighting the results from all individual rules into one

single output decision; 4) defuzzification of output to get

a value between 0 and 1. To develop a fuzzy logic con-

troller, two major components are required: 1) definition

of a membership function for each input/output parame-

ter; 2) designing the fuzzy rules. The membership func-

tion is a graphical representation of the magnitude of

participation of each input. The fuzzy logic rules use the

input membership values as weighting factors to deter-

mine their influence on the output sets [17].

Considering the number of malicious users is much

less than the total number of users, the median is less

vulnerable to the presence of the malicious users. So, the

filtering removes most of the malicious users.

Although filtering removes most malicious users,

some sensing results, which have been affected by fading

or by malicious users, might pass through filtering. The

performance of the system will degrade in these two

cases. To prevent this performance degradation, we pro-

pose to dedicate a trust factor (TF) to each user who

passed through filtering. The users, which have been

affected by fading or are malicious, are assigned with

lower trust factors. The users, which have a good channel

condition or aren’t suspicious to be malicious are as-

signed with higher trust factors. Trust factors are deter-

mined and normalized so that their summation becomes

one, i.e.

For several reasons, fuzzy logic is very appropriate for

using on secure spectrum sensing. One reason is that

there is no clear boundary between normal and anomaly

users. The use of fuzziness of fuzzy logic helps to

smooth the abrupt separation of normality and abnormal-

ity. Another reason is the reduction of miss detection and

false alarm probabilities. In the next section, we present

details of the fuzzy logic that we use in our secure spec-

trum sensing algorithm.

3. Secure Spectrum Sensing Algorithm



11.

TF k





(3)

By filtering and removing sensing results of malicious

users, performance of the system can be compensated to

some extent. First, the sensing results are passed through

a pre-filter. In our proposed algorithm, those sensing

results that are numerically distant from the rest of the

results are not considered in the final decision. We pro-

pose to use the median (med[k]) and standard deviation

(







) of energy values of sensing results in the com-

putation of upper and lower bounds of the filtering, at

time instant k, as follows:

In this paper we propose to use fuzzy logic to dedicate

each user a Trust Factor. As mentioned before, the fuzzy

logic is composed of membership functions for each the

input/output variables and fuzzy values. We select spec-

trum sensing results as input parameters to the fuzzy

controller in order to detect maliciou s users. For an input

parameter, three Gaussian membership functions are

designed: 1) always no malicious users; 2) trusted users;

and 3) always yes malicious users. Figure 2 shows the

E. MOEENTAGHAVI ET AL.

510

Figure 2. Membership functions of an input parameter.

three Gaussian membership functions for an input pa-

rameter. The output parameter also has two Gaussian

membership functions distributed in the range [0, 1] as

shown in Figure 3. These two membership functions are

called Fuzzy Trust Level (FTL) and Fuzzy Suspicious

Level (FSL). After defining the input parameters, the

fuzzy logic rules are designed. These rules are written

depending on the knowledge of secure spectrum sensing.

We discuss these rules in the following:

1) If (the sensing result is Trusted user) then (output is

FTL). This rule presen ts the sensing result of this user is

an expected value, so this user is normal and participates

in final decision in the fusion center;

2) If (the sensing result is always no malicious user) or

(the sensing result is always yes malicious user) then

(output is FSL), this rule presents the sensing result of

this user is an unexpected value, so th is user is malicious

and takes no part in final decision.

The output of the system shows the trust and suspi-

cious levels of each user. Those users whose sensing

results are near the median of the sensing results are as-

signed with higher trust levels and other users whose

sensing results are far from the median are assigned with

higher suspicious levels in the fuzzy logic. After De-

fuzzification for each user, we have a value for FTL and

a value for FSL in the range [0, 1]. The FTL shows the

degree of being a normal user and the FSL shows how

much malicious the user might be. In final decision, to

eliminate the effects of malicious users, we propose to

combine each sensing result, considering its fuzzy trust

level (FTL). First, we normalize fuzzy trust level of user

n at time instant k as follows:









FTL

FTL ,

FTL











(4)

where N denotes the number of users which have been

Figure 3. Membership functions of the output.

passed through filtering. Then, the final decision is

computed using FTLs of all N users as follows:

 

1FTLee.

nn T









(5)

If the value obtained in th e left side of the above equ a-

tion is greater than a given threshold (eT), the fusion cen-

ter will announce the presence of the primary signal. In

our proposed algorithm, the suspicious users who are

malicious or their sensing results are affected by fading

are assigned with lower fuzzy trust levels, so their effects

on the final decision are insignificant.

To achieve a better performance, the sensing results of

each user over a given number of (say L) measurements

are considered to obtain the final fuzzy trust level for

each user. In the computation of a fuzzy trust level, we

assign higher weights to those FTLs which are closer to

the present time, k, i.e.:









FTLFTL .

kLl k













(6)

Finally, these weighted fuzzy trust levels are normal-

ized according to Equation (4).

By considering previous and present behaviors of each

user in computation of the final fuzzy trust level, the us-

ers which behave maliciously for a period of time and

behave normally the rest of time, are detected and as-

signed with lower Fuzzy Trust Levels.

4. Simulation Results

In this section, the proposed secure spectrum sensing

algorithm is evaluated by simulations. The basic pa-

rameters are fixed and considered as a group of N = 50

secondary users. The mean received SNR of the channel

between the primary user and each of secondary users is

2 dB. Independent and identically distributed small scale

E. MOEENTAGHAVI ET AL.

511

fading channels are considered between any SU and the

PU, and the path loss is neglected.

In Figure 4, we assume a cooperative system with 10

“Always No” malicious users, each giving a value indi-

cating the absence of the primary user. To evaluate our

secure sensing algorithm, we compare probability of

detection (Pd) and probability of false alarm (Pf) of our

algorithm and three other cases, which are: 1) coopera-

tive spectrum sensing with no malicious user; 2) spec-

trum sensing with malicious users with no suppression;

and 3) secure spectrum sensing proposed in [5]. From

Figure 4 we can see that using our secure sensing algo-

rithm, the “Always No” malicious users are assigned

with low fuzzy trust levels and can not affect the per-

formance of the system. So, the probability of detection

(Pd) of the system would be close to that of cooperative

sensing with no malicious user.

In Figure 5, unlike Figure 4, we consider a sensing

system in which 10 users always announce the presence

of the primary user to the fusion center. From Figure 5

we can see that our proposed scheme can nullify the ef-

fects of malicious users in the final decision and has bet-

ter performance compared to those of previous works.

It’s notable that in the secure sensing, which was pro-

Figure 4. Probability of detection and false alarm in adversarial environment with 10 “Always No” malicious users.

Figure 5. Probability of detection and false alarm in adversarial environment with 10 “Always Yes” malicious users.

E. MOEENTAGHAVI ET AL.

512

posed in [5], the fading channel is not considered. By

applying the fading effects to this scheme, due to its fil-

tering structure, some malicious users affect the final

decision.

In Figure 6, we observe the probability of detection

according to the number of “Always No” malicious users.

From this figure, we can see that our proposed algorithm

is more robust than traditional ones. Our algorithm is

robust until 50% of the secondary users become mali-

cious and has a better performance compared to that of

[5].

In Figure 7, we observe the probability of false alarm

with varying the number of “Always Yes” malicious

users. As shown in the figure, our algorithm with effec-

tive malicious user detection can achieve an acceptable

performance.

5. Conclusions

In this paper, a new cooperative secure spectru m sensing

05 10 15 20 25 30 35

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Number of mali ci ous users

P robability of det ec t i on

P roposed Al go rithm

No malicious detection

Existing method [5]

Figure 6. Probability of Detection with varying the number of “Always No” malicious users.

510 15 2025 30 35 40

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Number of mal ici ous users

P robability of F als e A larm

Proposed A l gorit hm

No malicious detect ion

Exis ti ng m ethod [5]

Figure 7. Probability of False Alarm with varying the number of “Always Yes” malicious users.

E. MOEENTAGHAVI ET AL.

513

algorithm for malicious user detection in cognitive radio

networks based on fuzzy logic was proposed. In our

proposed secure sensing algorithm, first a pre-filtering is

designed to prevent the users, whose sensing results are

far from the others, take effect on final decision. Then,

based on the results of the filter output, the fuzzy pa-

rameters are obtained, and then according to fuzzy pa-

rameters, a fuzzy trust level is assigned to each user. Fi-

nally, the sensing results are combined in the fusion cen-

ter based on their fuzzy trust levels. Simulation results

show that our proposed algorithm can significantly nul-

lify the effects of malicious users. Moreover, it can alle-

viate the effect of fading channels. Furthermore, the

complexity of our proposed algorithm is much lower

than those of existing ones. In future work, we will de-

velop our secure sp ectrum sensing algorithm for the case

of using cyclostationary detectors (rather than energy

detectors used in this paper) and for more complex sce-

narios.

6. Acknowledgements

The authors would like to thank the Iranian Institute of

information and Co mmunication Technology (th e former

ITRC) for the financial support of this work.

7. References

[1] S. Haykin, “Cognitive Radio: Brain-Empowered Wireless

Communications,” IEEE Journal on Selected Areas in

Communications, Vol. 23, No. 2, 2005, pp. 201-220.

doi:10.1109/JSAC.2004.839380

[2] S. M. Kay, “Fundamentals of Statistical Signal Process-

ing: Detection Theory,” Prentice Hall, Upper Saddle

River, 1998.

[3] H. V. Poor, “An Introduction to Signal Detection and

Estimation,” Springer-Verlag, New York, 1994.

[4] S. Enserink and D. Cochran, “A Cyclostationary Feature

Detector,” Proceedings of the 28th Asilomar Conference

on Signals, Systems, and Computers, Pacific Grove, 31

October-2 November 1994, pp. 806-810.

doi:10.1109/ACSSC.1994.471573

[5] P. Kaligineedi, M. Khabbazian and V. K. Bhargava, “Se-

cure Cooperative Sensing Techniques for Cognitive Radio

System,” IEEE International Conference on Communica-

tions, Beijing, 19-23 May 2008, pp. 3406-3410.

[6] B. Shen and K. S. Kwak, “Soft Combination Schemes for

Cooperative Spectrum Sensing in Cognitive Radio Net-

works,” ETRI Journal, Vol. 31, No. 3, 2009, pp. 263-270.

doi:10.4218/etrij.09.0108.0501

[7] D. Cabric, S. M. Mishra and R. W. Brodersen, “Imple-

mentation Issues in Spectrum Sensing for Cognitive Ra-

dios,” Proceedings of 38th Asilomar Conference on Sig-

nals, Systems, Computers, Pacific Grove, 7-10 November

2004, pp. 772-776.

[8] R. L. Chen, J.-M. Park and K. G. Bian, “Robust Distrib-

uted Spectrum Sensing in Cognitive Radio Networks,”

Proceedings of the 27th Annual IEEE Conference on

Computer Communications, Phoenix, 13-18 April 2008,

pp. 1876-1884.

[9] K. J. Peng and Z. H. Tsai, “A Distributed Spectrum Sens-

ing Scheme Based on Credibility and Evidence Theory in

Cognitive Radio,” Proceedings of the 17th IEEE Interna-

tional Symposium on Personal, Indoor and Mobile Radio

Communications, Helsinki, 11-14 September 2006, pp.

1-5. doi:10.1109/PIMRC.2006.254089

[10] N.-T. Nhan and I. Koo, “A Secure Distributed Spectrum

Sensing Scheme in Cognitive Radio,” Proceedings of the

Intelligent Computing 5th International Conference on

Emerging Intelligent Computing Technology and Appli-

cations, Vol. 5755, 2009, pp. 698-707.

[11] S. Y. Xu, Y. L. Shang and H. M. Wang, “Double Thresh-

olds Based Cooperative Spectrum Sensing against Un-

trusted Secondary Users in Cognitive Radio Networks,”

IEEE International Vehicular Technology Conference,

Barcelona, 26-29 April 2009, pp. 1-5.

[12] W. K. Wang, H. S. Li, Y. Sun and Z. Han, “Attack-Proof

Collaborative Spectrum Sensing in Cognitive Radio Net-

works,” IEEE Annual Conference on Information Sciences

and Systems, Baltimore, 18-20 March 2009, pp. 130-134.

[13] E. MoeenTaghavi and B. Abolhassani, “Trustworthy

Node Detection in Cognitive Radio in Hostile Environ-

ments,” International Conference on Communication and

Electronics Information, Vol. 2, 2011, pp. 258-262.

[14] P. Kaligineedi, M. Khabbazian and V. K. Bhargava,

“Malicious User Detection in a Cognitive Radio Coop-

erative Sensing System,” IEEE Transaction on Wireless

Communications, Vol. 9, No. 8, 2010, pp. 2488-2497.

[15] L. A. ZADEH, “Fuzzy Sets,” Information and Control,

Vol. 8, No. 3, 1965, pp. 338-353.

doi:10.1016/S0019-9958(65)90241-X

[16] L. A. ZADEH, “Fuzzy Algorith ms,” Information a nd Con-

trol, Vol. 12, 1968, pp. 94-102.

doi:10.1016/S0019-9958(68)90211-8

[17] W. El-Hajj, F. Aloul, Z. Trabelsi and N. Zaki, “On De-

tecting Port Scanning Using Fuzzy Based Intrusion De-

tection System,” International Conference on Wireless

Communications and Mobile Computing, Crete Island,

6-8 August 2008, pp. 105-110.