Detection of AIS Spoofing in Fishery Scenarios

Abstract

For the purpose of maritime safety, information, and surveillance, almost all sea-going vessels have to participate in the Automatic Identification System (AIS). This system serves as a cooperative VHF-radio exchange of navigational and ships' information. Since AIS broadcasts self-declared information, it is open to fraudulent misuse by users. Based on different approaches to classification of maritime vessels, i.e., Random Forest, Voting-2-of-3, Decision Tree, Fuzzy Rule, and $k$ Nearest Neighbor, this contribution addresses the question, up to which accuracy it is possible, to detect fishery vessels with spoofed AIS-type based only on ship's positional, motion, and dimensions' AIS-data. For this purpose, in real-life AIS datasets from early summer 2017 the classification results of AIS fishery type are evaluated and compared.