Satellite Constellation Signal-Based Framework for Detecting GPS Spoofers

Abstract

The invention developed by the University of Central Florida and the University of Alabama is a robust, deep neural network model, named GANSAT, which is a defense mechanism against adversarial attacks that can degrade the performance of Global Positioning System (GPS) applications. This model is based on the framework of generative adversarial networks (GANs) to detect unauthorized or spoofed GPS signals. The presented model may predict positions in a GPS-degraded/denied environment using the novel concept of GPS satellite constellation fingerprint; the neural network implicitly learns the individual satellite’s relative position to each other and to the receiver, based on its contribution to the received GPS signal. The proposed GANSAT framework yields approximately 99.5% accuracy for identifying and filtering out the spoofed GPS signals from real ones.

Benefit

Market Application

Publications

GANSAT: A GAN Based System for Detecting GPS Spoofer using SATellite Constellation Fingerprinting, MobiCom ’20: ACM 26th Annual International Conference on Mobile Computing and Networking, Sep 1–25, 2020, London, United Kingdom.