On the Relation Between OC-LSSVM and Likelihood Test for Physical Layer Authentication

Physical layer authentication (PLA) allows a verifier to decide whether a received message has been transmitted by a legitimate or a malicious user by processing a series of features extracted from the physical channel over which the message traveled. However, to design the authentication check, typically, no information about the attacker is provided. In turn, either the statistics or a dataset of features is available for the legitimate user’s channel. When the statistics are known, a well-known good solution is the likelihood test (LT) while when a dataset is available, machine learning (ML) techniques are used. This letter focuses on one-class least-squares support vector machine (OC-LSSVM), showing that, with suitable kernels, it operates as the LT at training convergence. Numerical results are provided using wireless and underwater acoustic channels.