Reports

Tracking Down Dhows – 2 Techniques  to Detect Smuggling

The Challenges of Making the Invisible Visible

Dhows, a lateen-rigged ship with one or two masts, are commonly used in the Indian Ocean and are often involved in smuggling arms from Iran to Yemen. 

Dhows are ubiquitous in the Red Sea, but are also known to operate in the Gulf of Aden and the Persian Gulf. These vessels are difficult to detect, as they do not transmit their location via AIS, only GPS. They are usually detected and intercepted based on information from human intelligence (HUMINT) sources, which is limited in scale and coverage, and heavy on resources.

The prevailing belief is that no commercially available technology can detect dhows. Maritime domain awareness (MDA) systems cannot track vessels with no AIS transmission, making dhows essentially invisible to such tools.  

Windward’s AI-powered behavioral-analytics risk model can detect the vessels that interact or are likely to interact with dhows and pinpoint where unmanned assets should be deployed to improve interception chances.

This case study demonstrates an operational planning process devised by Windward, concluding with the revelation of suspicious ship-to-ship (STS) activity in the port of Bosaso, Somalia. 

In addition to generating leads in the form of vessels potentially engaging in illicit STS activities with dhows, the investigation also illuminates certain trends and behavioral patterns of note – providing tactical and strategic actionable insights.

Windward’s Operational Planning and Prioritization Approach

Somalia’s territorial waters have long been a hub for illicit activity, from human trafficking to arms smuggling. A simple geo-fencing approach, where all slow-moving vessels inside Somalia’s waters are identified to detect potential illicit activity is taking place, would be time-consuming, resource-heavy, and ineffective. 

A search for commercial vessels conducting slow-speed activity (below three knots) in Somalia’s territorial waters between May 2023 and May 2024 reveals nearly 700 vessels dispersed along Somalia’s shore (image 1 below). This is too broad to understand where to deploy assets.

Somalia's Territorial Waters
Image 1: Slow-speed activity in Somalia’s territorial waters (May 9, 2023-May 9, 2024). Source: Windward’s AI platform.

To increase effectiveness, two steps can be taken. 

  1. Build a profile to narrow down results even further. 
  2. Apply Windward’s predictive behavior risk models, which can identify vessels that are more likely involved in smuggling events.

Using Windward’s unique Sequence Search capability, we built a sequence consisting of: 

  • Vessels that conducted dark activity inside an area of interest (in this case, the Persian Gulf)
  • After the dark activity ended (while still in the Persian Gulf), the vessel slowed down its speed inside Somalia’s territorial waters.

As image 2 shows, this sequence search yields 129 vessels – a far more manageable number of leads compared to the nearly 700 before, but still not sufficiently narrowed. 

Somalia 2
Image 2: Numbers of vessels that have turned off their AIS transponders in the Persian Gulf and then slowed their speed to below three knots inside Somalia’s territorial waters (May 9, 2023-May 9, 2024). Source: Windward’s AI platform.

It was time for Windward’s predictive AI models to come into play. The ability to identify vessels likely involved in illicit activity is built into Windward’s system. By refining the search even further, to only show vessels marked for high risk of smuggling (of a variety of available risk types), we moved from 700 vessels to only a handful – 15 vessels, to be exact.

The search also reveals where most of the nefarious activity is taking place inside Somalia’s territorial waters. Based on the results (as seen in image 3 below), we can immediately deduce that the Horn of Africa, close to the port of Bosaso, is where most of the illicit activity likely takes place.

With this information, planning the mission becomes data-driven, with a clear and defined target area for deployment and further examination.

Somalia 3
Image 3: The number of high-risk vessels suspected of smuggling that have turned off their AIS transponders in the Persian Gulf and then slowed their speed below three knots inside Somalia’s territorial waters (May 9, 2023-May 9, 2024). Source: Windward’s AI platform.

Windward’s Approach for Investigating Leads with Tipping and Cueing

As mentioned above, Windward’s Sequence Search, combined with its predictive risk models, significantly narrowed the list of 700 vessels to 15 quality leads – making investigations feasible. To demonstrate, let’s pick one of the 15 vessels, a Palau-flagged cargo vessel.

15 vessels
Image 4: An example of a vessel suspected of smuggling that has turned off its AIS tracking system, later conducting a suspicious drifting pattern in Somalia’s territorial waters (April 23, 2024). Source: Windward’s AI platform.

As image 4 above shows, the vessel went dark inside the Persian Gulf for 430 hours (nearly 19 days). Its prolonged period of dark activity was followed by slow-speed activity inside Somalia’s territorial waters. 

For context, it is important to look at the vessel’s behavioral patterns before this sequence of activities. Using Windward’s Maritime AI™ platform, we can examine the vessel’s entire voyage. On February 20, 2024, it turned off its transponder after leaving the UAE, beginning a dark activity period of 17 days. This was followed by port calls in Pakistan and India, and, finally, a stop in Bosaso. 

Except for turning off its AIS transmission outside the Persian Gulf, the vessel did not seem to be behaving in a way that raised suspicion. But the benefit of this approach is that it can be used to prompt a tipping and cueing process. To gain more insight into the vessel’s behavior, we can bring satellite images to an area where we know the vessel conducted an unusual loitering event. This is to see if the ship met other vessels during this time:

Satellite 1
Image 5: An electro-optical (EO) imagery overlay, focused on the area where the vessel has conducted unusual loitering (April 23, 2024). Source: Windward’s AI platform.

Looking at the next image (image 6, below), we see two vessels. The first, our target vessel, is transmitting. But there is also a smaller, non-transmitting ship next to it: a traditional dhow, a 40-meter-long vessel. 

Satellite 2
Image 6: An EO imagery overlay, focused on the area where the vessel conducted unusual loitering, shows a ship-to-ship transfer between the suspected target and a 50-meter-long, non-transmitting vessel (April 23, 2024). Source: Windward’s AI platform.

It is not possible to know for certain that these vessels were engaged in contraband smuggling during their meeting. Yet the meeting between two vessels, one non-transmitting and the other at high risk for smuggling after it had just conducted a suspicious pattern of activities, is certainly reason enough to deploy assets for further examination. 

One Vessel of MANY…

The Palau-flagged cargo vessel shown above is one of several general cargo vessels engaging in STS meetings with non-transmitting vessels appearing to be dhows. 

Based on the sequence of activities detailed above, we identified numerous vessels involved in suspicious journeys from the Persian Gulf to Bossaso, Somalia, a port traditionally used for smuggling between Somalia and Yemen. To detect STSs with traditional dhows, we searched for events where vessels slowed below three knots in the port’s vicinity.

Five examples are below:

50m vessel 1
Image 7: A Togo-flagged cargo vessel meeting with a 50-meter-long, non-transmitting vessel (November 6, 2023). Source: Windward’s AI platform.
50m vessel 2
Image 8: A Palau-flagged cargo vessel meeting with a 50-meter-long, non-transmitting vessel (March 30, 2024). Source: Windward’s AI platform.
50 m vessel non-emitting
Image 9: A Palau-flagged cargo vessel meeting with a 50-meter-long, non-emitting vessel (February 3, 2024). Source: Windward’s AI platform.
Satellite
Image 10: A Palau-flagged cargo vessel meeting with a 50-meter-long, non-emitting vessel (January 25, 2024). Source: Windward’s AI platform.
Map
Image 11: A Palau-flagged cargo vessel meeting with a non-transmitting vessel in Iran (April 20, 2024) and in Somalia (May 8, 2024). Source: Windward’s AI platform.

An analysis of this trend reveals the following:

  • Using Sequence Search to go back a year (May 2023-May 2024), we only began to see this pattern of activity emerging after October 7, 2023 – with the first event taking place in November 2023. 
  • The vessels engaging in this pattern of behavior are general cargo vessels flying flags of convenience (except for one vessel Togo-flagged vessel, all other vessels fly the Palau flag). They either turn off their AIS tracking system or conduct course deviations in the Persian Gulf/Iran before engaging in ship-to-ship operations with traditional dhows near Bosaso. 
  • Most vessels detected conducting ship-to-ship operations near Bosaso with non-transmitting vessels, likely dhows, simply sailed from the port of Kandala in India to Bossaso. We identified three variants of this pattern:
    • 1. Ships turning off their AIS in the Persian Gulf for a long time (possibly loading cargo during a dark STS), then conducting slow-speed activity near the Bosaso port in Somalia, where they engaged in an STS with a non-transmitting dhow.
    • 2. Ships conducting a route deviation, another indication of a dark STS operation in the Persian Gulf. Image 11 (above) shows one vessel conducting a dark STS meeting in Iran following a route deviation (April 20, 2024), possibly to load, and then another dark STS meeting in Somalia (May 8, 2024), likely to unload.
    • 3. Ships sailing from Kandla port in India followed by slow-speed activity in Somalia’s territorial waters.

Windward Can Help!

It is a misconception to believe there is no technological solution to detect dhows. While dhows may be “invisible,” the ships they are meeting are not. Instead of tracking dhows, an endless and resource-intensive endeavor that is impossible to scale up, it is possible to reverse-engineer known patterns to generate leads for further exploration and investigation

The traditional approach of simple geofencing is not sufficient. Combining AI technologies for quality lead generation with human expertise for lead investigation is essential.

Windward can help with the tipping and cueing process for unmanned systems and autonomous assets. For this report, we used satellite imagery to investigate the leads, but this could easily be conducted via a drone, an aerial unmanned asset, etc.

Contact us for more information.

I Want to Detect Dhows