A data driven approach to maritime anomaly detection

Abstract

The operational community has long identified anomaly detection systems as vital for increasing the effectiveness of maritime surveillance systems, since the huge quantities of data produced today quickly reduce their effectiveness in the field. The limited range of currently available maritime anomaly detection systems rely heavily on expert knowledge and hardcoded rules; thus, limiting their scope and effectiveness only to known situations and patterns of behaviour. By contrast, data driven systems learn from the data itself and can thus generalise well to new tasks and previously unseen situations. In this paper, we present an overview of the anomaly detection system developed in the context of the European Commission H2020 funded project BigDataOcean, and specifically the “Maritime Security and Anomaly Detection” pilot. Through a combination of unsupervised machine learning methods and behavioural analytics, the developed system is capable of i) automatically modelling shipping routes at a global scale; ii) constructing vessel specific profiles and class baselines; and iii) detecting deviations from patterns of normalcy in real time.

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 732310.