Detecting Dark Ship-to-Ship Transfers Using Deep Learning and Satellite Imagery

Core Concepts

This study pioneers the application of deep learning techniques for the identification of dark Ship-to-Ship Transfers (STS) in satellite imagery, addressing a crucial gap in maritime surveillance.

Abstract

The study leverages a combination of Automatic Identification System (AIS) data and PlanetScope satellite imagery to detect dark STS events, where vessels manipulate or disable their AIS for illicit purposes. The results reveal over 400 dark transshipment events in the Kerch Strait between 2022 and September 2023, underscoring the prevalence of such deceptive maritime activities.
The methodology involves three key steps:

Identifying STS in AIS data by linking together ships that spend extended periods in close proximity.
Training a YOLOv8m object detection model to accurately classify different types of cargo vessels and STS events in satellite imagery.
Cross-referencing the satellite-based STS detections with AIS data to identify "dark" transshipment events where fewer than two AIS signals are present.

The demonstrated efficacy of this approach signifies its potential for enhancing maritime security and facilitating timely intervention in illicit shipping practices.

Stats

The study utilizes over 1 million unique vessel positions from AIS data spanning January 2021 to September 2023, as well as PlanetScope satellite imagery collected between January 2017 and September 2023 over the Kerch Strait.
The trained YOLOv8m model achieved an overall F1 accuracy of 97% and an mAP50 of 99%, indicating a high degree of accuracy in identifying the six classes of vessels in satellite imagery.

Quotes

"Following Russia's invasion of Ukraine, G7 countries instituted a price cap of $60 per barrel for Russian crude oil, meaning that Vladimir Monomakh would be carrying roughly $49 million worth of oil when at capacity. However, on the date of the transfer visible in Figure 6 the world price of brent crude was $78 per barrel. As such, circumventing the price cap would increase the sale price of the ship's cargo by around $14 million."

Key Insights Distilled From

Automatic Detection of Dark Ship-to-Ship Transfers using Deep Learning and Satellite Imagery

by Ollie Ballin... at arxiv.org 04-12-2024

https://arxiv.org/pdf/2404.07607.pdf

Deeper Inquiries

How can the proposed methodology be expanded to cover a wider geographical area beyond the Kerch Strait?

To expand the proposed methodology to cover a wider geographical area, several steps can be taken. Firstly, increasing the dataset by incorporating AIS data and satellite imagery from other regions would provide a more diverse training set for the deep learning model. This would involve collecting AIS data and satellite imagery from different maritime zones to ensure the model's generalizability. Additionally, enhancing the object detection model's capabilities to recognize different types of vessels and transshipment patterns commonly found in various regions would be crucial. Moreover, collaborating with international maritime organizations to access a broader range of AIS data and satellite imagery could help in scaling the methodology to cover multiple maritime hotspots globally.

What counter-measures could be employed by vessels engaged in illicit transshipment to evade detection by this system?

Vessels engaged in illicit transshipment may employ various counter-measures to evade detection by the system proposed in the study. One common tactic could be the manipulation of AIS data, such as spoofing or turning off AIS transponders to avoid being tracked. Additionally, vessels might engage in complex maneuvering patterns to make it challenging for satellite imagery to capture their activities accurately. Another strategy could involve using signal jamming devices to disrupt satellite communication and prevent the transmission of real-time vessel positions. Furthermore, vessels might operate in congested maritime areas or near legitimate shipping routes to blend in with regular maritime traffic and avoid suspicion.

What broader implications might the prevalence of dark ship-to-ship transfers have on global trade, energy markets, and geopolitical tensions?

The prevalence of dark ship-to-ship transfers can have significant implications on global trade, energy markets, and geopolitical tensions. In terms of global trade, illicit transshipment activities can distort market prices, facilitate the smuggling of contraband goods, and undermine fair competition among legitimate traders. In energy markets, dark ship-to-ship transfers involving oil tankers can lead to price manipulation, tax evasion, and regulatory non-compliance, impacting the stability of energy markets worldwide. Geopolitically, the use of dark transshipment for sanctions evasion or illicit activities can strain international relations, trigger diplomatic disputes, and escalate tensions between countries. Moreover, the unchecked proliferation of dark ship-to-ship transfers can weaken maritime security, disrupt supply chains, and pose challenges to global efforts to combat illegal activities at sea.