Cooperative Spectrum Sensing Algorithm Based on Support Vector Machine against SSDF Attack

Abstract

In cognitive radio networks (CRNs), a spectrum-sensing-data-falsification (SSDF) attack is a security issue. To deal with SSDF attacks, a support vector machine (SVM) based scheme is proposed here. Our scheme analyzes secondary user's behaviors from multi-round records of energy values, and obtains a novel evaluation index, classification accuracy. In particular, the concepts of recognition probability and misclassification probability are introduced, and the tradeoff relationship between misclassification probability and threshold of classification accuracy is theoretically obtained. Moreover, the asymptotic optimal property is derived. It enables excellent adaptability for malicious secondary user (MSU) detection in various scenarios. Exhaustive simulation results demonstrate that our proposed scheme outperforms other existing approaches.