SSDF Attack And Defense Strategies In Cooperative Spectrum Sensing Of Cognitive Radio Networks

With the rapid development of wireless communication technologies, the contradiction between the increasing demand of the spectrum band and the limited spectrum resource has been becoming one of the main factors restricting the development of wireless communications. Cognitive radio (CR) is emerged as a promising technology to resolve the spectrum scarcity, in which the spectrum sensing, as one of the most important parts to realize the CR, has become a hot research topic. In order to improve the performance of spectrum sensing, cooperative spectrum sensing (CSS) is introduced, in which many CR users collaborate to find the vacant spectrum. However, the CSS introduces new security threats into the cognitive radio network (CRN), where the most typical attack is the spectrum sensing data falsification (SSDF) attack. In the SSDF attack, malicious CR users intentionally send falsified sensing results to fusion center (FC), which results in either the excessive interference on the primary user or preventing CR users from using the detected vacant spectrum band.This dissertation focuses on the SSDF attack problems and defense strategies in the CSS of CRNs. Specifically, the SSDF attack problem in the CSS with soft fusion, the SSDF attack problem in the CSS with M-ary quantized data fusion, and the joint spectrum sensing and resource allocation (JSSRA) problem under SSDF attack are studied.The main contributions of this dissertation are summarized as follows.First, aiming at the SSDF attack problem in the CSS based on the linear data fusion, the conditions to nullify the detection capability of the FC are derived. For two probabilistic SSDF attack models in the CSS with soft data fusion, their corresponding conditions for nullifying the detection capability of the FC are analyzed by setting the modified deflection coefficient of the FC to be zero. Moreover, a probabilistic SSDF attack model in the CSS with M-ary quantized data fusion is defined, and the corresponding blind condition is derived.Second, in order to resolve the SSDF attack problem in the CSS with soft data fusion, a reputation-based CSS scheme in the CRN is proposed. To mitigate the effect of the SSDF attack, a reputation adjusting method based on the improved ROCQ reputation model and the corresponding reputation-based CSS scheme are proposed. In the proposed reputation adjusting method, FC adjusts each CR user’s reputation according to its sensing result, sensing accuracy and sensing credibility. Moreover, in the proposed reputation-based CSS scheme, the reputation degree of each CR user is used as the weight coefficient of the linear data fusion at the FC. Simulation results show that the proposed reputation-based linear CSS scheme can eliminate the harmful effect of the SSDF attack.Third, in order to resolve the SSDF attack problem in the CSS with M-ary quantized data fusion, an SSDF attack defense strategy is proposed. A malicious CR user identification method based on the characteristic of the maximum output entropy (MOE) quantization method in the CSS scheme with M-ary quantized data is presented. FC estimates the probability mass functions (PMFs) of each CR users’ quantized data. By utilizing the character that PMFs of each quantization level with the MOE quantization method are constant, malicious CR users are identified and excluded from the data fusion process, In addition, we present an identification threshold selection method, and prove the effectiveness and convergence of the proposed malicious CR user identification method. Numerical results show that, with the given quantization method, the proposed method can remove malicious CR users successfully without any information about the attacker’s strategy, and achieve a performance improvement in the CSS with M-ary quantized data fusion under SSDF attack.Fourth, since the resource allocation process can be affected by the CSS process disturbed by the SSDF attack, we investigate JSSRA problem with SSDF attack. The JSSRA problem is formulated as a weighted-proportional-fairness resource allocation problem, where the weighted coefficient is set to be the trust degree of each CR user. The optimization problem is decoupled and solved to obtain the sensing time, the number of cooperative CR users, and the allocated data transmission time. As the number of cooperative CR users is given, a reinforcement-learning-based cooperative user selection method which selects proper CR users for the CSS based on their sensing performance and trust degrees is proposed. Numerical results show that the proposed JSSRA scheme deals with the SSDF attack well in cooperative sensing process to improve the system robustness, and achieves a significant system utility gain in resource allocation process.