Research On Key Issues For Spectrum Sensing In Cognitive Radio Networks

To resolve the contradiction that spectrum resources are scarce and most of licensed spectrum bands are idle in most of time, cognitive radio which is a technology based on dynamic spectrum access has been widely researched. Spectrum sensing paves the way for realization and practical application of cognitive radio. Terminal equipment for cognitive radio works on the licensed bands only if no primary users work on the bands. In this dissertation, the performance for cooperative spectrum sensing in cognitive radio networks are analyzed, and strategies for different kinds of spectrum sensing data falsification attacks are proposed. The main work and contributions are presented as follows:1. Analyzing operation mode for hard decision and soft decision in cooperative spectrum sensing, obtaining the threshold for quantization bits when the performance for soft decision outperforms that for hard decision. This threshold should be used as a criterion when soft decision is utilized by fusion center. For the purpose of minimizing the global error probability, analyzing and obtaining the optimal fusion rule for hard decision when the sensing channel is fading and the reporting channel is imperfect.2. When malicious users launch uniform attacks to cognitive radio networks, countermeasures to attacks for hard decision and soft decision are proposed respectively. Firstly, when there exist trusted nodes in the network, a scheme is proposed to counter attacks which malicious users launch independently or cooperatively. Secondly, the network is attacked by smart attacks, and there are no trusted nodes, a scheme based on reputation of each node in the network is proposed. The reputation value for each node in the network is updated according to the reputation area which the node belongs to. The proposed method associating with OGK algorithm mitigates the effect of malicious users, maintaining the robustness of cooperative spectrum sensing. Thirdly, when the local sensing technologies utilized by secondary users have some uncertainties, deficiencies of evidence theory for cooperative spectrum sensing are analyzed in this dissertation, and a scheme based on the evidence similarity between nodes in the network is proposed. The proposed scheme obtains the evidence similarity and utilizes the similarity matrix to get the evidence support rating of each node from other nodes in the network, then discards the malicious nodes from the data fusion process.3. The non-uniform attacks are divided into static non-uniform attacks and dynamic non-uniform attacks. A scheme to counter static non-uniform attacks is proposed according to the properties of the information theory for each node. Meanwhile, a scheme is proposed to counter dynamic non-uniform attacks through deploying k-medoids algorithm to distinguish malicious users. These schemes enhance the detection performance and maintain the robustness when the percentage of malicious users in the network increases.4. In the situation that there is no fusion center in the network, drawbacks of the scheme for cooperative spectrum sensing based on consensus are analyzed when there exist malicious users in the network. A method is proposed to deteriorate the attacking success probability of malicious users through analyzing the data trend of a node in the last three loops, reducing the probability of successful attacking effectively.5. When there are some selfish nodes in the network, each node tends to save energy to transmit data and decrease the probability of participating in the spectrum sensing. The scenario that there exist more primary users is analyzed, and the evolutionary game theory is utilized to decide a secondary user whether participates in spectrum sensing and which licensed channel to sense. The proposed scheme maximizes the utility of each node through sensing game in the cognitive radio network.