Relay selection scheme based on machine learning for enhancing the physical layer secrecy in cognitive radio networks

In this paper, a machine learning schemes are proposed to enhance the physical layer security in cognitive radio network in the presence of an eavesdropper and primary user. Firstly, we apply the support vector machine (SVM) based scheme as a machine learning based scheme to classify and select a relay node to assist the secondary user transmitter (SU-Tx) and maximize the secrecy rate and satisfy the interference constraint at PU. Then, we develop a deep neural network (DNN) based scheme to classify and select the best relay to assist the SU-Tx. Compared to the conventional optimal selection (OS) scheme, we prove that the proposed DNN-based scheme can achieve the same secrecy performance and the proposed scheme can substantially reduce the feedback overhead. Moreover, the proposed scheme based on SVM can achieve a good performance with small complexity compared to DNN and conventional OS. In contrast, the conventional OS requires knowledge of the eavesdropper's channel, which is impractical, whereas the proposed scheme based on DNN and SVM do not assume knowledge of the eavesdropper's channel so our proposed scheme is less complex with a small feedback overhead, e.g., at least 58% feedback overhead could be reduced.