Machine Learning-based Physical Layer Authentication using Neighborhood Component Analysis in MIMO Wireless Communications

This paper proposes a machine learning (ML)-based physical layer authentication scheme in the multi-input multi-output (MIMO) wireless communication environment. In the proposed scheme, a neighborhood component analysis(NCA) based feature selection method is imployed to improve authentication performance. In particular, NCA based selection mitigates performance degradation even in the worst case scenario, where an intruder replaces the legitimate transmitter by using the feature selection. The selected features are classified using a radial basis function (RBF) kernel-based support vector machine (SVM). Through simulation results, it is shown that the proposed scheme obtains better authentication performance compared to the conventional approaches. Furthermore, it can achieve the average of area under the receiver operating characteristic curve of 0.9965 or higher even for worst cases of authentication scenarios.