Behave Well: How to Win a Pop Vacant Band via Cooperative Spectrum Sensing

Abstract

Cooperative spectrum sensing (CSS) for vacant licensed bands is one of the key techniques in cognitive radio networks. However, current CSS schemes focus on ensuring an efficient cooperation among secondary users (SUs), but ignoring their competition. At the same time when several SUs want to a vacant band, how to win this pop vacant band for an SU becomes more and more important. Inspired by the idea that an SU who always behaves well will win a pop vacant band more easily, we propose a competition scheme called BehaveWell (BW) in this paper. By analyzing the main threats against CSS, competitive coefficient is introduced to evaluate each SU's past behaviors in CSS. A higher competitive coefficient is very helpful for an SU to win a pop vacant band. This BW scheme can not only enhance a healthy competition among SUs, but also improve the security of CSS. Simulations verify the effectiveness of the proposed scheme.

1. Introduction

With the rapid development of wireless communication technologies and the huge demand of the capacity for wireless applications, the wireless frequency spectrum has become increasingly scarce. However, a large portion of the assigned spectrum bands are not utilized efficiently. According to the Federal Communications Commission (FCC), temporal and geographical variations in the utilization of the assigned spectrum range from 15% to 85% [1]. To solve the contradiction between the spectrum scarcity and low spectrum utilization, cognitive radio have been considered as a useful technology, which allow the licensed users to share their vacant bands with unlicensed users who are not assigned bands, thereby increasing the efficiency of the spectrum utilization [2]. The licensed users are also called the primary users (PUs) and the unlicensed users are the secondary users (SUs) in cognitive radio networks.

Cooperative spectrum sensing (CSS) is a key to the opportunistic use of assigned spectrum bands in cognitive radio network, since it enables SUs to find the vacant bands in the case of deep shadowing and multipath fading. The main idea of CSS is to ensure the sensing performance by exploiting spatial diversity via the observations of spatially located SUs [3]. By cooperation, CSS can eliminate the negative effect from deep shadowing and multipath fading by sharing sensing data and make a reliable cooperative decision.

As always, we ignore an issue that the number of SUs is far more than PUs in cognitive radio network. So, the vacant bands are very scarce, and it is often the case that a few SUs compete for a vacant band at the same time. In this case, how to win this pop vacant band via CSS becomes more and more important. On the other hand, CSS is being compromised by malicious threats. False sensing, one of the most famous threats, is launched by malicious SUs to deceive other SUs to acquire a wrong cooperative decision via CSS.Many efforts have been made to suppress this threat, such as outlier detection [4], shadow-fading correlation [5], expectation maximization [6], goodness-of-fit [7], trust or reputation [8-11], etc. However, these methods fail to address selfish sensing threat. There may exist some selfish SUs who are unwilling to cooperate but enjoy sensing data from others. Such threat may seriously degrade the performance of CSS and even make a well designed CSS scheme useless [12]. To suppress selfish threat, recent efforts are mainly paid to game theory [13-15]. Additionally, SUs must never interfere with PUs in CSS [2]. Some SUs may long occupy vacant bands even though PUs come back, which is prone to causing harmful interference to PUs. Currently, false sensing, selfish sensing and long occupation are the main threats against CSS, but not a method can suppress them together with a scheme. From the perspective of competition, the SUs who launch these threats in CSS will be punished to obtain a lower competitive coefficient which can reduce their opportunities or even make them have no chances to get any vacant band, whereas the SUs behave well in CSS will have a chance to get a vacant band.

In this paper, we propose a competition scheme called BW to enhance a healthy competition among SUs. As we know, each SU plays two roles in the CSS environment, the role of cooperating SU reporting sensing data and the role of initiator SU enjoying sensing data. Noticing this, our design idea is that an SU can win a pop vacant band more easilier than others if he always behaves well while playing the role of cooperating SU. The main contributions of this paper are as follows.

The organization of this paper is as follows. In section2, preliminaries related on CSS are described. We construct the BW scheme in section 3 and describe its implementation strategies in section 4. Simulation analysis of BW is given in section 5. Finally, we conclude this paper in section 6.

 

2. Preliminaries

The CSS process can be viewed as a parallel fusion network [16]. As shown in Fig. 1, a central authority called fusion center (FC) controls the process of CSS: individual sensing, data reporting and data fusion [3].

Fig. 1.Modeling CSS as a parallel fusion network.

In the CSS process, licensed channel is the selected licensed frequency band where a physical point-to-point link between the PU transmitter and each SU for observing the vacant band, and control channel is a physical point-to-point link between each SU and FC for sending individual sensing data [3]. It can be seen that the two types of channels are given by cognitive radio network. Thus, the CSS process among SUs seems not waste any more spectrums.

Additionally, the CSS process will work when at least one SU send a CSS request to FC since these SUs cannot distinguish a vacant band from a deep shadowing effect and multipath fading. However, at the same time when several SUs want to the vacant band of a PU, how FC choose an SU to use this pop vacant band. In this paper, our main task is to help FC to make a choose decision that the SU who behave well in the CSS process will get a pop vacant band more easily.

 

3. Proposed Competition Scheme

In this section, we first describe some related definitions of the BW scheme, and then present an individual scheme to evaluate individual competitive coefficients by analyzing three threats in CSS. Finally, a general evaluation framework is designed to incorporate them as a single value.

3.1 Related Definitions

Some definitions are introduced in the BW scheme. To understand the design of our scheme better, we describe these new definitions in advance.

3.2 Evaluation of Individual Competitive Coefficients

From the perspective of competition, our scheme is designed to inspire SUs behave well. That is, if an SU rarely launches threats against CSS, he will have more chances to get a vacant band. Currently, there are three threats that are mainly damaging the performance of CSS.

In summary, it can be seen that the SUs who launch these threats in CSS behave binary (i.e., positive or negative). Such binary behaviors can be described as

Recently, one of the most popular designs using binary input to evaluate a variable ranging from 0 to 1 is based on beta function. It first counts the number of positive and negative behaviors a user has conducted, and then calculates the variable with beta probability density functions (PDF) denoted by [18].

where θ is the probability of sensing behaviors, 0≤θ≤1, α>0, β>0.

Without loss of generality, let T={t1, …, tk,…, tn} denote the set of threats, where tk is the k-th threat. The k-th individual competitive coefficient (cik) of an SU (Take SUi as an example) corresponding to tk can be evaluated with beta function as: cik=Beta(posik+1, negik+1). posik and negik denote the number of positive sensing and negative sensing about the k-th threat launched by SUi. Without any prior observations, posik=negik=0 and hence, cik=Beta(1, 1).

Consider the case Γ(x)=(x-1)! when x is an integer [19]. It can be deduced that the expectation value of the beta function is given by: E[Beta(α, β)]=α/(α+β). Thus, cik can be further described as

In our scheme, we marke the CSS threats {FS, SS, LO} as {t1, t2, t3}. Then (posi1, negi1) represent the number of real and false sensing respectively, ci1 is the individual competitive coefficient of SUi corresponding to FS. Similarily, ci2 and ci3 denote the individual competitive coefficient corresponding to SS and LO respectively, which can be evaluated by (posi2, negi2) and (posi3, negi3) successively.

3.3 General Evaluation Framework

From the perspective of suppressing threats, we can evaluate individual competitive coefficients for each SU. Then, a general evaluation framework is necessary to incorporate these coefficients as a single value. In our scheme, the competitive coefficient (Ci) of SUi corresponding to T can be evaluated as

To promote a healthy competition in cognitive radio network, f(·) should satisfy three requirements.

Based on the above requirements, we consider the following scenarios {prior, common, zero} for evaluating Ci.

➢ Prior scenario

For each cik>δ, it means SUi behaved well in the past. FC should give SUi a prior authority to use a pop vacant band. His competitive coefficient is evaluated as

For a good expandability, new individual competitive coefficient must meet 0≤cik≤1.

➢ Common scenario

For some cik>δ, it means SUi behaves well sometime. That is, this SU has launched some malicious threats before. FC should give SUi a common authority. His competitive coefficient is evaluated as

It can be seen that the malicious behaviors will decay Ci severely, result in increasing his difficulty to win a vacant band. Of course, an opportunity should be offered to correct his misconduct. For instance, if SUi doesn’t launch the k-th threat during a period of time, his cjk value will be increased by 0.1. But, once he is found to launch this threat again, more severe punishment will be given to him, such as his cik value will be decreased by 0.2 immediately.

➢ Zero scenario

For each cik<δ, it means SUi always behaves maliciously and launches all threats. FC has to give SUi a zero authority. His competitive coefficient is evaluated as

This SU is hopeless, and FC or even cognitive radio system should reject to his request for any vacant band. He can get nothing from CSS.

 

4. Implementation Strategies

The effectiveness of supporting competitive coefficient in CSS depends not only on the factors and metric for evaluating competitive coefficient, but also on the implementation of the BW scheme. Typical issues in implementing BW in cognitive radio network include "pop band allocation" and "competitive coefficient maintain for mobility environment".

4.1 Pop Band Allocation

When have known that competitive coefficient is used in the case when there are several SUs want to win a pop vacant band at the same time, such as time h. Procedure 1 is designed to describe how SUs can win a pop vacant band by their competitive coefficients. When several SUs send requests to ask a pop vacant band at time h, two issues should be addressed before performing Procedure 1. 1) They have different competitive coefficient (i.e. each Ci ≠Ck for i≠k) at time h. 2) They have the same competitive coefficient (i.e. each Ci = Ck for i≠k) at time h, how FC decides who will pop up?

For the first issue, it is easy for FC responds to these requests from the highest to the lowest according to their competitive coefficients. To address the second issue, three priority strategies can be introduced to help FC, including Good Behaviors Priority, TTL Priority and Random Priority.

➢ Good Behaviors Priority

For some cik>δ in the set T={t1, …, tk,…, tn}, as we know, that means SUi has launched some malicious threats before. Let mi denote the number of malicious threats launched by SUi. Specially, mi =0 when each cik>δ. Then, FC can compare the m values for different SUs to decide who will pop up. Take SUx and SUy as an example, if mx

➢ TTL Priority

In the BW scheme, TTLi (time-to-live) represents the life time of SUi from who joined cognitive radio network to time h. Obviously, TTLi =h-τi ( the initial time when SUi joined cognitive radio network). When several SUs have the same competitive coefficient at time h, the SU with highest TTL would pop up. For an SU with a smaller TTL, he is likely to have fewer opportunities to launch malicious threats. Especially for TTLi =0, that means SUi is a newcomer who has done nothing in the network and his competitive coefficient is 1. But, once he launches malicious threats in the future, his competitive coefficient will be decayed sharply by Eq.(5). For an SU with a higher TTL, he is likely to have more opportunities since he can maintain a higher competitive coefficient for a long time. Therefore, we should give a chance to the SU with highest TTL when several SUs have the same competitive coefficient at time h. This strategy is named as tp() in Procedure 1.

➢ Random Priority

Initially, FC chooses an SU randomly to use the pop vacant band when several SUs have the same competitive coefficient at time h. But such choice is restricted by another index r which is the number of the SU who has been chosen randomly in the Random Priority strategy. For SUi, ri =0 if he has never been chosen and ri will be added by 1 when he is chosen. Then, FC can compare the r values for different SUs to decide who will pop up. Of course, it is possible that there are more than two SUs with the same low r value. In this case, FC should chooses an SU randomly from them again. This strategy is named as rp() in Procedure 1.

Let S={SU1, …, SUi,…, SUn} denotes the set of several SUs who send requests to ask a pop vacant band at time h, U is the set filtered from S and |U| resprents the amount of elements in U.

4.2 Competitive Coefficient Maintain for Mobility Environment

We have known that the process of CSS is in the charge of FC which is usually a base station. However, the management area of FC is local and the sensing distance of an SU is limited. Take Fig. 2 as an example, local areas compose the cognitive radio network. In order to facilitate the depiction of this figure, we draw a PU in each local area. In practice, there are a FC and several PUs coexiting with SUs in a local area. If PUs are fixed devices, they will not move. If not, they may move from a local area to another one.

Fig.2.Example of local areas in cognitive radio network.

From Fig. 2 we can find a question that competitive coefficient is usually utilized in a local area. Can SUi’s competitive coefficient in area 1 be used in area 2 when he moves from area 1 to area 2? To address this question, E3C is introduced to maintain competitive coefficient for SUs in different local areas. E3C records the indexes {(posi1, negi1), (posi2, negi2) , (posi1, negi2) τi, ri} for SUi. To make E3C work better, four characteristics should be abided by E3C.

Assuming SUi, SUj denote cooperating SU and initiator SU in Procedure 2. As shown in Fig. 1, di is the individual sensing data from SUi and d is the cooperative decision from FC. To ensure E3C update effectively, detection method of the CSS threats {FS, SS, LO} by FC should be considered in Procedure 2.

 

5. Simulation Analysis

We would perform five simulations to validate the BW scheme and show its effectiveness.

5.1 Simulation Setup

The simulations are performed based on the energy detection, in which the primary signal is a baseband QPSK modulated signal under the AWGN (additive white Gaussian noise) environment. The general simulation setup is shown in Table 1.

Table 1.Description of simulation elements

In the simulations, the SUs are split into two types: malicious SUs and honest SUs. The percentage of malicious SUs is set to 40%. The behavior pattern for them is to launch three threats in CSS, while honest SUs always behave well.

All simulations are executed by cycle-based fashion. At each cycle, all SUs are selected to perform CSS actions with each other randomly. After a few cycles, a competitive network topology is gradually formed by the BW scheme. FC then uses this scheme to perform CSS actions at each cycle, and update the competitive coefficients on the corresponding SUs.

5.2 Simulation Results

In the first simulation, we choose a malicious and honest SU randomly to observe the updating process of their competitive coefficient. An SU’s competitive coefficient is associated with his past sensing behaviors. As shown in Fig. 3, an honest SU’s competitive coefficient is greater than δ and tends to 0.9 after 20 cycles, since he always behaves well in CSS. Conversely, a malicious SU’s competitive coefficient is rarely larger than δ. We can also find that BW makes the malicious SU’s competitive coefficient fluctuates with the increase of cycles. This is because the common and zero scenarios can suppress the boost of competitive coefficient against threats. With a lower competitive coefficient, a malicious SU is difficult to win a vacant band.

Fig. 3.Updating process of competitive coefficients.

We then perform three simulations to validate the robust of BW in terms of its suppressing three threats.

In the second simulation, we analyze the receiver operating characteristic (ROC) curves, the relationship between the probabilities of detection (Qd) and false alarms (Qf), which is usually to validate the sensing performance of CSS. As shown in Fig. 4, we can see that the ROC curves of the “AND” and “Majority” rule with BW are better than traditional rules without protection, which indicates that the BW scheme under the FS threat can enhance the performance of the two fusion rules significantly after filtering out malicious SUs. When one SU reports “1” in the “OR” rule, the PU signal is considered to be present. It also can be seen that malicious SUs have a little influence on the “OR” rule by analyzing the ROC curves.

Fig. 4.ROC curves under the FS threat.

To analyze the effectiveness of BW under the SS threat, we observe the total number of query and sensing at each cycle in the third simulation. As shown in Fig. 5, BW can increase the total number of sensing to a certain extent under the SS threat. In the BW scheme, if an SU (such as SUi) provides nothing in CSS for a long time, the growing number of negi2 will decay his competitive coefficient. To main a high competitive coefficient, BW can inspire SUs to provide sensing data.

Fig. 5.Total number of query and respond at each cycle under the SS threat.

In the fourth simulation, we analyze the effectiveness of BW under the LO threat. BW makes some SUs scruple that lots of complaint will cause the decrease of their competitive coefficient. Therefore, BW can suppress the ratio of complaint at each cycle under the LO threat, as shown in Fig. 6.

Fig. 6.Suppressing ratio of complaint at each cycle under the LO threat.

In addition, some opportunistic spectrum access (OSA) [20-21] technologies also involve the issue to coordinate SUs to utilize a pop vacant band to aviod their collision. Nevertheless, they neglect that the security would affect the fairness of allocating a pop vacant band. That is, some SUs launch threats to maximize their benefit, but no punitive measures are adopted to prevent them from utilizing vacant bands. In fact, the actions that malicious SUs launch threats in CSS will generate a large number of malicious responses which is mainly caused by the FS, SS and LO threat in CSS. If not suppressed, the number of malicious responses would become more and more terrible with the increase of cycles. So, the best method to compare BW with OSA is to test how they can suppress malicious responses. The fifth simulation is performed to observe this comparison. As shown in Fig. 7, BW can suppress the growth of malicious responses significantly compared with OSA with the increase of cycles. In the BW scheme, the SUs who launch threats are punished in the competition of vacant bands. To increase their opportunities for vacant bands, some malicious SUs would become to behave well.

Fig. 7.Suppressing malicious responses.

 

6. Conclusion

In this paper, we have proposed a competition scheme called BW to enhance a healthy competition among SUs. The key idea of our scheme is that an SU who always behaves well will win a pop vacant band more easily. Inspired by this, we analyze the main threats against CSS, and thus introduce the competitive coefficient which is associated with each SU’s past sensing behaviors in CSS. A higher competitive coefficient can contribute significantly to winning a pop vacant band for an SU. The implementation strategies of BW are also described in detail. Simulation results show that BW can make malicious SUs obtain a low competitive coefficient and improve the robust of CSS.