Detecting Illegal, Unreported, and Unregulated Fishing through AIS Data and Machine Learning Approaches

Illegal, unreported, and unregulated (IUU) fishing is a critical issue impacting the health of the marine ecosystem and global security. It is linked to crimes such as human trafficking and destroys local economies through over-fishing, collapsing species populations and throwing the ecosystem out of balance. A major contributor to this is the deceptive use of Automatic Identification System (AIS) beacons on fishing nets, buoys, or lines, to conceal illegal fishing hauls from local vessels and enforcement authorities. Enhancing the capacity to recognize fraudulent AIS devices, and predicting their operational areas of activity, is crucial for regulatory agencies aiming to reform the current state of fishing practices. This paper focuses on analyzing geospatial AIS data from Southeast Asia and provides three machine learning modelling approaches to aid in the detection of IUU fishing occurrences. We utilize semi-supervised classification, unsupervised clustering, and standard neural network models as approaches to identify irregular AIS beacons, suspected to be IUU fishing instances, based on device patterns and characteristics. For further exploration, we also conducted regional analysis of IUU fishing features to identify areas of suspected heightened illegal activity. Our results show that movement and positional features of AIS devices can serve as successful indicators, alongside machine learning techniques, for IUU detection. This research serves as foundational analysis to improve the field of IUU fishing using modelling techniques.