Behavioral Model Anomaly Detection in Automatic Identification Systems (AIS)

Over 90% of all goods in the world, at some point in their life, are carried on a vessel at sea. Currently, the maritime industry relies on the Automatic Identification System (AIS) for collision avoidance, vessel tracking, and vessel awareness while operating at sea. AIS is a plaintext, unencrypted, unauthenticated protocol and, as such, is vulnerable to various types of attacks. Malicious actors can alter the AIS location of a vessel by spoofing a vessel or alter the channel the AIS receiver is using to send nefarious information to a vessel privately. With the advent of the Ocean of Things (OoT), vessels are sharing more information than vessel location alone at sea. As this information becomes critical for safe and efficient operation at sea, we in this work present a novel approach of applying machine learning to build behavior models for vessels at sea. These models allow vessels to detect anomalous communication from vessels nearby, thus enable vessels to determine the quality of the messages shared between each other and, more critically, identify malicious vessels' behaviors.