Mitigating Autonomous Vehicle GPS Spoofing Attacks through Scene Text Observations

Abstract

This paper investigates both from an empirical and a systems-based perspective, how surrounding textual information can be leveraged towards the mitigation of Autonomous Vehicle (AV) and self-driving cars Global Positioning System (GPS) signal spoofing attacks. The paper presents and proposes methods of how AVs and self-driving cars can extract, as they travel along a trajectory, surrounding textual information through machine-learning based Scene Text Recognition (STR). The paper researches and proposes geospatial models which can be applied to the extracted textual information in order to build a text-based geolocation system for the purposes of validating the received GPS signal. The ultimate contribution of the paper is to lay the groundwork towards enhancing the Cybersecurity of the current and future Autonomous Vehicle and self-driving car ecosystem by addressing its Achilles heel, namely insecure and inaccurate geolocation due to GPS spoofing attacks.