Extreme Learning Machine Based Identification of Malicious Users for Secure Cooperative Spectrum Sensing in Cognitive Radio Networks

Cognitive radio (CR) technology has evolved over the traditional radio to successfully utilize the unused frequency spectrum. In CR the secondary users (SUs) perform cooperative spectrum sensing to access the available frequency band. The opportunistic nature of sensing prevents any interference with primary users (PUs) in the network. However, the presence of security threats like malicious users (MUs) strongly influences the performance. In CR network, MUs act like normal SUs and transmit false information to the fusion center and degrades the performance. To overcome this issue, we proposed an extreme learning machine (ELM) based approach to classify the legitimate SUs with the MUs. In this work, ELM is used as a classifier to separate the legitimate SUs and MUs. Extensive simulation results are presented to highlight the effectiveness of the proposed approach. The proposed approach highlights significant improvement in terms of training time and provides better trade-off compare to the other competitive techniques in the literature.