AI-Based Runtime Monitoring for Detecting Geoposition Falsification with Normalizing Flows

Ensuring the integrity and authenticity of geoposition data in Cooperative Awareness Messages (CAMs) is crucial for the safety and reliability of Vehicle-to-Everything (V2X) communication systems. Falsified or manipulated geoposition data can lead to severe safety risks and undermine trust in these systems. This paper introduces a novel runtime monitoring approach that leverages Normalizing Flow models to detect geoposition falsification and identify misbehavior in CAM transmissions. Unlike traditional rule-based or heuristic approaches, the proposed method utilizes a Normalizing Flow model to implicitly learn the spatial dynamics, operational range, and coverage areas of Roadside Units (RSUs) based on authentic CAM data. By modeling the spatial distribution patterns of legitimate transmissions, the system can assess the likelihood of incoming geoposition data in real time. Anomalies that significantly deviate from learned patterns are flagged as potential falsifications or evidence of misbehaving transmitters. The approach is evaluated in the Test Area Autonomous Driving Baden-Württemberg (TAF-BW), a realworld testbed for autonomous driving technologies, demonstrating its potential. ... mehrBy leveraging Normalizing Flows to monitor CAM transmissions in real-time, our method offers a robust solution to mitigate geoposition falsification attacks. This not only contributes to safer and more reliable V2X systems, but also paves the way for future research in advanced security measures for intelligent transportation systems.