The Return of Dazzle: How WWI Camouflage Found a Second Life in Ukraine's Drone War

A century after dazzle camouflage confounded German U-boat targeting in the First World War, Russian forces are repainting military trucks in stark black-and-white patterns to defeat the AI-driven targeting systems now deployed against them.

By Monexus Staff Writereastern europe6-minute read31 May 2026☆ Save ↗ Share ⎙ Print

In the autumn of 1917, British Admiral到手ty commissioned a small army of artists — including the marine painter Norman Wilkinson — to devise a new kind of ship camouflage. Unlike the muted grays meant to blend with open water, these patterns were loud: angular blocks of navy blue, lime green, and chalk white arranged in zigzags and spirals across hulls and superstructures. The goal was not concealment. It was confusion. At distances and angles that U-boat periscopes could not predict, a dazzle-painted ship was supposed to look like it was moving faster, or in a different direction, than it actually was — making torpedo calculations nearly impossible.

The Admiralty called it "dazzle camouflage." The sailors called it "Razzle Dazzle." And for a few years, hundreds of vessels steamed across the Atlantic in stripes that looked more like Kandinsky paintings than naval vessels.

A century later, the concept has returned — not on warships but on supply trucks rumbling across the flat steppe of eastern Ukraine.

On 30 May 2026, open-source intelligence monitors and military bloggers reported that Russian forces had begun painting military trucks in bold black-and-white geometric patterns, explicitly designed to defeat the AI-assisted targeting systems that Ukrainian drones use to identify and strike vehicles on the front lines. The stripes, sources confirmed, employ the same logic as their WWI predecessors: not hiding the object, but making it computationally harder to classify.

The Pattern Meets the Machine

Ukrainian First Person View (FPV) drones — the cheap, wire-guided quadcopters that have become one of the defining weapons of the current conflict — rely heavily on machine vision algorithms to track moving targets. Operators feed the drone a visual template of a vehicle; the onboard software steers toward the closest match. The system works fast and at scale, but it is trained on datasets — real imagery of trucks, tanks, and supply vehicles — that reflect the visible spectrum as it has historically presented itself.

A truck painted in high-contrast black-and-white disruptively may fall outside the statistical confidence range the model expects. The vehicle is still visible. It is still moving. But the algorithm tasked with classifying it as "military truck" encounters ambiguity where it expected certainty. That ambiguity costs fractions of a second in targeting lock — fractions that, in a moving vehicle on an open road, can be the difference between a successful strike and a miss.

Ukrainian drones also use thermal and infrared imaging, which complicates the countermeasure. Heat signatures do not respect surface color. But analysts note that dazzle patterns applied to engine compartments and exhaust vents can create thermal irregularities that further confuse automated tracking — disrupting the gradient the software expects to see across a vehicle's silhouette.

The reported resurgence of dazzle on Russian supply vehicles is not the first time the pattern has surfaced in the current conflict. Throughout 2023 and 2024, OSINT analysts documented irregular camouflage on Russian vehicles near the contact line, though early instances were attributed to field improvisation rather than systematic adoption. The reports from late May 2026 suggest a more deliberate rollout, with vehicles photographed near occupied Donetsk Oblast showing consistent geometric coverage rather than ad hoc weathering.

A History Written in Stripes

The original dazzle scheme emerged from a specific technical problem: submarines attacking from the beam, at relatively short range, needed to estimate a ship's course and speed in seconds. Traditional camouflage, optimized for hiding from distant lookouts, was useless against a periscope peer checking the horizon at eye level. A ship that looked like it was turning when it was going straight, or sailing fast when it was stationary, might cause a torpedo to burst behind the target rather than alongside it.

Norman Wilkinson's original proposal, submitted to the Admiralty in early 1917, was explicitly titled "Dazzle Painting of Ships." His argument rested on psychology and geometry: the human eye, and by extension the human judgment that translated visual cues into targeting data, could be overwhelmed by patterns that refused to resolve into a stable shape. Over the next two years, some 2,000 merchant vessels and warships received dazzle treatments. Postwar analysis suggested the scheme was at best partially effective — German torpedo doctrine evolved faster than camouflage theory — but the psychological impact on crews was real, and the visual legacy proved durable.

The pattern reappeared in naval contexts during the Second World War, was periodically revived in maritime design, and in recent years has been adopted by some naval architects for anti-piracy applications and by automakers referencing its aesthetic. But its re-entry into a live military context — this time against algorithmic targeting rather than human periscope operators — represents something close to a direct continuity with the original problem. The target has changed; the strategy has not.

The Geometry of Uncertainty

What makes dazzle effective against machine vision is not fundamentally different from what made it effective against a submarine commander: both systems are prediction engines. A human observer calculated likely trajectories; a targeting algorithm calculates likely class. The vulnerability in both cases is the same: a model, whether biological or computational, relies on priors — expectations about what a target looks like and how it behaves. Dazzle disrupts those priors.

The high-contrast geometric patterns are particularly effective against convolutional neural networks — the class of machine learning models most commonly used in object detection and classification — because those models are sensitive to texture and edge regularity. A smooth-painted truck has coherent edges across its body panels; a dazzle-treated truck has edges that break and re-form at every stripe boundary, effectively creating hundreds of small apparent objects where the algorithm expects one. The vehicle becomes, from the software's perspective, a densely cluttered scene rather than a discrete target.

This is not a permanent solution. Machine learning models can be retrained; Ukrainian drone operators can be given imagery of dazzle-painted vehicles to supplement their training datasets. Within weeks of a new countermeasure's deployment, the opposing system adapts. The arms race between camouflage and detection has been running for a century; the current iteration just runs faster.

What Comes Next

Whether Russian forces are systematically painting their supply convoys in dazzle patterns — or whether the late May sightings represent a localized initiative by individual units improvising solutions to a real problem — is not yet confirmed by Ukrainian military briefings or Western defense analysts. Open-source researchers caution that vehicle camouflage in contested areas is notoriously difficult to assess from imagery alone; units modify their equipment for a range of reasons, and a pattern photographed in one location does not indicate a coordinated doctrine.

If the practice is expanding, however, it signals something specific about the nature of the current battlefield: the conflict is increasingly shaped by the interaction between human operators and machine decision-support systems on both sides. The truck on the road is not merely a logistical asset — it is a data point in a targeting algorithm, an image in a training set, a node in a model that has learned what Russian military vehicles look like from thousands of hours of footage.

That framing — of soldiers responding to the machine's expectations rather than the enemy's — would be recognizable to the young officers who oversaw Atlantic convoys in 1918. The technology has changed; the geometry of confusion has not. Dazzle survives because it exploits something fundamental about how prediction systems, human or artificial, process the world. And as long as those systems govern the targeting of weapons, someone, somewhere, will be painting stripes on a truck to make them wrong.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

https://t.me/ClashReport
https://en.wikipedia.org/wiki/Dazzle_camouflage
https://en.wikipedia.org/wiki/Norman_Wilkinson_(artist)