A comprehensive analysis and novel methods for on-purpose AIS switch-off detection

Abstract

Maritime security, as well as monitoring of illegal and illicit activities such as smug- gling and Illegal, Unreported and Unregulated (or IUU) Fishing are reliant on Au- tomatic Identification System (AIS). As AIS is a self-reporting system, determining where and when suspect activities are taking place is difficult when vessels are simply able to decide to switch off their AIS device on purpose. The AIS on-purpose switch- off problem attempts to differentiate between typical loss of signal between a vessel and base station with an intentional shut down of the on-board AIS transponder. Previous works, while few, have attempted various strategies for solving this prob- lem. From statistical analysis, to machine learning, to reconstructing message signal strength from base stations. All of these works, however, have some area in which they lack. In this work we provide a comprehensive analysis and propose improvements to what is considered the state-of-the-art approach to solving this problem. The im- provements on the previous works’ extrapolation algorithm showed a statistically significant increase to the F1-Score ranging from 0.03 to 0.051 at an α of 0.05 as well as an almost halving of the standard deviation. Digging deeper into the results of the various parameters and the relationship between their values and the overall performance provides even more promising results, showing exactly where it is that the original baseline fails and how the newer approaches proposed here are able to compensate for those original deficiencies.