GNSS Anti-spoofing Detection based on Gaussian Mixture Model Machine Learning
Zejian Feng, Chee Kiat Seow, Qi Cao
Abstract
Nowadays, the security of Global Navigation Satellite System (GNSS) has raised much more concerns due to the reliance on its position, velocity, and timing (PVT) information, which is of vital importance to various Internet of Things (IoT) systems, robotics, 5G technology and many applications of Intelligent Transportation Systems (ITSC). It has been shown that GNSS system can be easily spoofed and masqueraded to provide ill intent payload damages. This paper proposes a novel algorithm based on unsupervised machine learning Gaussian Mixture Models (GMM) to provide anti-spoofing capability of GNSS signal such as GPS signal. It segregates GPS signals that are not under spoofing, from spoofed GPS signals that will result in malicious changes of pseudo-range measurements. It has been found out that the proposed GMM clustering algorithm is able to cluster the positions generated by the un-spoofed GPS signals properly and return the PRN (pseudo-range noise) codes of the satellites without spoofing effectively. The proposed GMM clustering algorithm could cluster the position points generated by non-spoofed signals properly by more than 90% and 77% accuracy for one and three spoofed satellites respectively.