GNSS Signal Classification based on Machine Learning Methods

Nowadays, spoofing attacks are increasing due to the widespread usage of navigation and positioning systems. Another reason for increasing these attacks is the appearance and easy accessibility of low-cost software-defined radios (SDR). Such attacks need to be properly detected and, if possible, mitigated in time. Therefore, it is necessary to develop efficient systems for spoofing detection. There are many different spoofing detection methods which include hardware or software even machine learning methods. In this paper, detection and classification of spoofed Global navigation satellite system (GNSS) signals is performed by using different supervised machine learning methods and specific features. Two different datasets are used, including publicly available SatGrid dataset, which is not used in the examined and available articles as far to authors’ knowledge, and our adapted and modified dataset. Verification of our dataset, tested in the controlled indoor environment, is performed. Obtained results show that all used ML models classify signals with high accuracy. Two methods Support Vector Machine (SVM) and Neural Networks (NN) exhibit slightly better accuracy compared to other methods.