Transfer Learning based GPS Spoofing Detection for Cellular-Connected UAVs

Abstract

Unmanned Aerial Vehicles (UAVs) are set to become an integral part of 5G and beyond systems with the promise of assisting cellular communications and enabling advanced applications and services, such as public safety, caching, and virtual/mixed reality-based remote inspection. However, safe and secure navigation of UAVs is a key requisite for their integration in the airspace. The GPS spoofing is one of the major security threats to remotely and autonomously controlled UAVs. In this paper, we propose a machine learning-based, mobile network-assisted UAV monitoring and control system that allows live monitoring of UAVs' locations and intelligent detection of spoofed positions. We introduce the Convolutional Neural Network (CNN) in the edge UAV Flight Controller (UFC) to locate a UAV and detect any GPS spoofing by comparing differences between the theoretical path loss computed by UFC and the corresponding path loss reported by the connected base station (BS). To reduce the detection latency as well as to increase the detection accuracy, transfer learning is leveraged to transfer the CNN knowledge between edge servers when the UAV handovers from one BS to another. The performance evaluation shows that the proposed solution can successfully detect spoofed GPS positions with an accuracy rate above 88% using only one BS.