Physical-Layer Authentication Based on Extreme Learning Machine

Most physical-layer authentication techniques use hypothesis tests to compare the radio channel information with the channel record of Alice to detect spoofer Eve in wireless networks. However, the test threshold in the hypothesis test is not always available, especially in dynamic networks. In this letter, we propose a physical-layer authentication scheme based on extreme learning machine that exploit multi-dimensional characters of radio channels and use the training data generated from the spoofing model to improve the spoofing detection accuracy. Simulation results show that our proposed technique can significantly improve the authentication accuracy compared with the state-of-the-art method.