Physical Layer Authentication with Likelihood Test Using Machine Learning with Artificial Dataset

In physical layer authentication (PLA) mechanisms, a verifier applies a test to decide whether a received message has been transmitted by a legitimate user or an intruder, according to some measured channel features (CFs). When the legitimate CF statistics are known, a well-known good solution is the likelihood test (LT). When a dataset of legitimate CFs is available, machine learning (ML) models can be used to perform one-class classification (OCC). Still, currently, i) while a good understanding of how ML models make decisions is important to ensure security, these models are not explainable and ii) statisticsand ML-based solutions appear as distinct solutions. In this paper, we aim at bridging such a gap by obtaining ML PLA verifiers that operate as the LT via neural network (NN) and least-square support vector machine (SVM) models, trained as two-class classifiers on the single-class dataset and an artificial dataset for the negative class. The artificial dataset is obtained by generating CF vectors uniformly distributed over the domain of the legitimate class dataset. In turn, we show that autoencoder classifier generally does not provide the LT. Numerical results are obtained considering PLA on both wireless and underwater acoustic channels.