Collaborative Spectrum Sensing Method Based on Optimized SVM and Credibility Weighted Fusion

This paper proposes a novel method to resist Spectrum Sensing Data Falsification (SSDF) attacks using the optimized Support Vector Machine (SVM). The proposed method involves the optimization of SVM kernel parameters, the recognition of malicious secondary users (MSUs) through feature vectors analysis, and the fusion of sensing data utilizing credibility weighting. The optimized SVM model is employed for the detection of MSUs, thereby improving detection probability and enhancing spectrum sensing performance in complex sensing environments. Besides, the fusion of sensing data, with credibility serving as a weight, strengthens the system's resistance against attacks. Simulation results demonstrate the effectiveness of the proposed method in effectively defending against SSDF's attacks, particularly under Report False Attack scenarios, while outperforming existing methods in terms of spectrum sensing performance.