Attack Detection And Robustness Algorithm Based On Collaborative Sensing In Cognitive Radio Networks

Cognitive radio(CR) is an intelligent technology to improve spectrum utilization and gets more and more considerable attentions. As one of the key issues in cognitive radio, collaborative spectrum sensing technology is able to improve sensing performance and increase access opportunity for under-utilized spectrum while avoiding interference to the primary users. However, it also opens an opportunity for certain secondary users(defined as attackers) for their attack objective, in which they corrupt the decision-making process by sending false sensing information in collaborative group, to achieve the pre-occupied opportunities of spectrum channels. This kind of attack is called spectrum sensing data falsification(SSDF), which can not only disrupt the performance of collaborative sensing, but also probably decrease the availability of spectrum. Therefore, it is significant to design an effective attack detection algorithm against SSDF attack for the security in cognitive radio networks.In order to improve the availability of collaborative sensing and maximize the spectrum efficiency, this thesis proposes a reputation-based collaborative sensing attack detection algorithm. Firstly, based on the reliability, the algorithm labels with a reputation each secondary user in each cluster of collaborative group, and updates their sensing reports by using a difference value function with reputation-weighted, so that to obtain a updated value represented as the local sensing decision of the certain cluster,which is quite close to the real status of primary users. Then, the global sensing decision is obtained by the fusion center based on the weighted-calculation optimization on all collected local sensing decisions of clusters. Finally, the reputation values of all secondary users are updated based on their previous sensing reports, and the honest users and attackers are identified by fusion center. Simulation results show that the proposed algorithm can obtain an updated sensing report value that is closer to the real status of primary user, and improve the detection accuracy of primary users. Moreover,the collaborative sensing performance of the proposed algorithm is more stable in hostile environments with several attackers, which considerable guarantees higher spectrum efficiency.Considering the certain exchange model of sensing reports of all secondary users in collaborative group, and in terms of the attack detection efficiency as well as detectionrobustness, this thesis further proposes a common control channel-based(CCC) security collaborative attack detection algorithm. Firstly, this algorithm establishes a conman control channel to refine the collaborative sensing model, and the attackers are detected by the algorithm with associated reputations. Next, with the internal punishment function(IPF) and direct search(DS) method, the algorithm calculates the optimized thresholds under the situation where a better attack detection efficiency is obtained.Finally, the robustness of the algorithm is verified by employing “Kullback-Leibler Diversity”(KLD) method. Simulation results show that the proposed algorithm not only can achieve a better efficiency on attack detection, but also maintain a considerable robustness performance, which has profound significance to address the SSDF attack in collaborative spectrum sensing and guarantee the security in cognitive radio networks.