Adaptive Physical Layer Authentication Using Machine Learning With Antenna Diversity

The heterogeneous characteristics of wireless mobile networks within the Internet of things (IoT) create authentication challenges due to the large number of devices with diverse requirements and capabilities. Physical layer authentication (PLA) can provide solutions for this heterogeneous environment using wireless channel attributes. In the paper, an adaptive lightweight physical layer authentication (ALPLA) scheme is proposed using machine learning (ML). Antenna diversity at the receiver is exploited to increase the number of features to achieve a high authentication rate (AR). A one-class classifier support vector machine (OCC-SVM) is used with the magnitude and real and imaginary parts of the received signal at each receive antenna as features. One-class classification is a ML technique for outlier and anomaly detection which uses only legitimate training data. The sounding reference signal (SRS) in the 5G uplink radio frame is employed to obtain the features. The proposed scheme is evaluated in an urban environment under different mobility conditions. Results are presented which show that this scheme provides a high AR with sufficient antenna diversity. Further it is superior to other approaches in the literature.