Contextual verification for false alarm reduction in maritime anomaly detection

Automated vessel anomaly detection is immensely important for preventing and reducing illegal activities (e.g., drug dealing, human trafficking, etc.) and for effective emergency response and rescue in a country's territorial waters. A major limitation of previously proposed vessel anomaly detection techniques is the high rate of false alarms as these methods mainly consider vessel kinematic information which is generally obtained from AIS data. In many cases, an anomalous vessel in terms of kinematic data can be completely normal and legitimate if the "context" at the location and time (e.g., weather and sea conditions) of the vessel is factored in. In this paper, we propose a novel anomalous vessel detection framework that utilizes such contextual information to reduce false alarms through "contextual verification". We evaluate our proposed framework for vessel anomaly detection using massive amount of real-life AIS data sets obtained from U.S. Coast Guard. Though our study and developed prototype is based on the maritime domain the basic idea of using contextual information through "contextual verification" to filter false alarms can be applied to other domains as well.