Support Vector Machine Process Against Probabilistic Byzantine Attack for Cooperative Spectrum Sensing in CRNs

In view of the spectrum shortage problem of wireless devices and applications, cognitive radio (CR) allows secondary users (SUs) with a signal detection function to opportunistically access spectrum resources being authorized to primary users (PUs) by means of cooperative spectrum sensing (CSS). However, the inherent openness of CR technology provides an opportunity for malicious SUs (MSUs) to launch Byzantine attack, therefore undermining the premise of CR. For this reason, motivated by advantages of machine learning, we make use of support vector machine (SVM) to identify MSUs by providing a maximum margin hyperplane in this paper, in which the generated spectrum sensing data features benefit from the PU status in the training process. This paper makes an in-depth analysis on the SUs’ sensing results in the presence of a large-scale probabilistic Byzantine attack by the SVM process and provides a method linking with a reputation structure to identify those potential MSUs and mitigate the negative impact of Byzantine attack on CSS. Finally, a series of numerical simulation result shows that the security of the CSS process can be guaranteed by selecting those reliable sensing results identified by means of SVM, especially in a large-scale attack.