Are the days of art forgers finally numbered? A new artificial intelligence (AI) tool might just provide the answer. Art historians and physicists at Case Western Reserve University (CWRU) in Ohio have joined forces and trained a computer system to reportedly identify the unique brushstrokes of artists with 96% accuracy.

“The idea was that analysing the brushstroke was going to create a fingerprint,” explained Kenneth Singer, a physics professor and lead researcher at CWRU. “We found that even at the brush bristle level, there was a fair level of success in sorting the attribution. Frankly we don’t really understand that, it’s kind of mind boggling actually when you think about it, how the paint coming off a single bristle is indicative of what we’re calling the artist’s unintentional style.”

Published in the journal Heritage Science, the research shows how these fingerprints could help art historians settle issues of attribution and recognise forgeries. The researchers even hope to develop “unbiased and quantitative methods” to identify different hands when a painting was made by several artists or produced in a workshop.

“I wouldn’t say it’s foolproof; I’m a scientist. But I would say it’s a powerful tool,” said Singer.

To find the key to unlocking the fingerprints, students at Cleveland Institute of Art were asked to paint copies of a photograph of a water lily. Three-dimensional topographical scans of the paintings’ surfaces were made with a profilometer. They were then divided into tiny squares – some as small as half a millimeter – and analysed by convolutional neural networks (CNNs) to identify “unintentional style”.

Previous research in the field has largely analysed high-resolution images of paintings; researchers at Rutgers University published a study in 2017 that photographed more than 80,000 individual strokes in 300 drawings by artists like Henri Matisse (1869-1954), Pablo Picasso (1881-1973), and Egon Schiele (1890-1918). But, for the very first time, researchers at CWRU have analysed the painted surfaces of the canvases themselves.

The research team is now collaborating with conservation firm Factum Arte to analyse ‘Portrait of Juan Pardo de Tavera’, painted in 1609 by El Greco (1541-1614). As well as identifying different hands, the AI has located before unknown areas of restoration, which occurred after the painting suffering extensive damage during the Spanish Civil War.

In the future, the AI could be harnessed to analyse paintings with even less surface texture like watercolours or drawings. “We’re at the point where we’ve just figured out the basics of a concept and our first attempt ended up being spectacularly successful beyond our wildest dreams,’’ reflected Elizabeth Bolman, chairman of the art history department at CWRU. “Where this goes from here, we can all dream.”