Detection of spoofed AIS: Simulated tracks vs. real maritime data

Abstract

The Automatic Identification System (AIS) is a valuable tool for enhancing maritime safety and security, primarily through its vessel tracking and collision avoidance functions. However, AIS is vulnerable to various cybersecurity threats, with simulated spoofed AIS tracks emerging as a significant concern. This paper analyses the stochastic kinematics of multiple vessels recorded in the Black Sea. Additionally, several Machine Learning models are evaluated for their effectiveness in distinguishing between genuine and spoofed maritime tracks. Accuracies exceeding 98% were obtained. The main concept of this study arises from the recognition that predicting future trajectories of maritime ships is susceptible to measurement and process errors. Measurement errors are primarily induced by inaccuracies in Global Navigation Satellite Systems, while process errors stem from factors such as weather conditions, wind, currents, and inconsistencies in vessel steering. Mathematical models typically generate spoofed tracks that lack the error variations observed in genuine data when predicting future positions, velocities, and headings. By analyzing and understanding these sources of error, this study demonstrates the potential to distinguish genuine maritime trajectories from simulated ones, thereby enhancing the detection of spoofed AIS tracks.

Keywords

• maritime anomaly detection
• ais spoofing
• simulated trajectories
• maritime surveillance