Tracking Shadow Vessels in the Arctic with AI

The Arctic is becoming increasingly difficult to monitor. As geopolitical tensions rise and shipping routes expand, nearly half of so-called "shadow vessels" are deliberately turning off or spoofing their identification systems to operate undetected. Traditional tracking relies on AIS (Automatic Identification System) transponders—but what happens when ships simply turn them off?

That's the problem we set out to solve with Arctic Overwatch.

Finding Ships by Their Wakes

Our approach was inspired by a simple insight: every ship leaves a wake, and that wake is unique. Like a fingerprint on water, these patterns can't be faked or disabled. Using Synthetic Aperture Radar (SAR) satellite imagery, we could see these wakes clearly, even in challenging Arctic conditions.

The challenge was teaching a computer to recognize and classify them.

Building the Detection System

We built a convolutional neural network (CNN) trained on 458 SAR images of vessels and their wakes. Rather than focusing on the ships themselves—which can be obscured or misidentified—our model learned to analyze the wake patterns trailing behind them.

The processing pipeline works in four stages: cleaning up noisy satellite data, detecting wake patterns with our CNN, generating a unique fingerprint for each wake, and validating it against available AIS records.

Why It Matters

Arctic Overwatch demonstrates that we don't need to rely solely on cooperative tracking systems. By combining satellite imagery with deep learning, we can maintain surveillance even when vessels try to disappear. This approach is scalable too—critical for monitoring the vast, remote expanses of Arctic waters where traditional patrol methods aren't practical.

The system isn't just about catching bad actors. It's about bringing transparency and accountability to one of the planet's most vulnerable and strategically important regions.