A data-driven method for falsified vehicle trajectory identification by anomaly detection

The vehicle-to-infrastructure (V2I) communications enable a wide range of new applications, which bring prominent benefits to the transportation system. However, malicious attackers can potentially launch falsified data attacks against V2I applications jeopardize traffic operation. To ensure brought by it is critical protect from those cyber-attacks. existing literature on defense solution that protects very limited. This paper aims fill this research gap proposing data-driven method identify trajectories generated compromised connected vehicles (CVs). A trajectory embedding model, inspired word model natural language processing (NLP) community, developed. proposed generates vector representations vehicle be used compute similarities between trajectories. consists two steps. In first step, historical are train neural network and obtain second step computes distance matrix each pair identifies using hierarchical clustering algorithm. Simulation experiments show has high detection rate (>97.0%) under different attack goals with varying CV penetration rates 100% 25%, while false alarm remains low. It great potential implemented in trajectory-based such as state estimation signal control, safeguard system cyber threats.