Physical Layer Authentication for Satellite Communication Systems Using Machine Learning

The vertical heterogeneous network (VHetNets) architecture aims to provide global connectivity for a variety of services by combining terrestrial, aerial, and space networks. Satellites complement cellular networks to overcome coverage and reliability limitations. However, the services of low-earth orbit (LEO) satellites are vulnerable to spoofing attacks. Physical layer authentication (PLA) can provide robust satellite authentication using machine learning (ML) with physical attributes. In this paper, an adaptive PLA scheme is proposed using Doppler frequency shift (DS) and received power (RP) features with a one-class classification support vector machine (OCC-SVM). One class-classification is a ML technique for outlier and anomaly detection which uses only legitimate satellite training data. This scheme is evaluated for fixed satellite services (FSS) and mobile satellite services (MSS) at different altitudes. Results are presented which show that the proposed scheme provides a higher authentication rate (AR) using DS and RP features simultaneously compared to other approaches in the literature.