AI smashes the idea that all fingerprints are unique – and could solve cold cases

Every fingerprint is unique, right? Well, computer says no.

Engineers at Columbia University have used artificial intelligence (AI) to smash one of the most basic and long-held beliefs in forensics. Kind of.

Their program found that while each finger does have its own unique pattern of swirls and loops, there are telltale giveaways within them matching one print to another on the same hand – a breakthrough that could help solve cold cases and even prove innocence in cases of a wrong conviction.

Imagine the scenario. Two crime scenes. At the first, only a single thumbprint was found. At the second, an index fingerprint was picked up. Until now, there was no way, using these prints alone, to connect one to the other.

But Columbia’s AI could change that.

The team, led by Columbia engineering undergraduate Gabe Guo, fed a government database of 60,000 fingerprints into the network in pairs. Some of the pairs belonged to the same person, some to two different people.

Over time, as the team continued to tweak the system, it became better and better at identifying which were pairs and which were not. Accuracy for a pair from the same individual hit 77%, and was significantly higher when from two people.

The team sent their results to a forensics journal, but the paper was rejected. The anonymous reviewer and editor concluded that ‘It is well known that every fingerprint is unique’, and so it would not be possible to detect similarities even if the fingerprints came from the same person.

Undeterred, Mr Guo and the team worked on making their AI even more accurate.

Fingerprints have been used in forensics for a long time (Picture: Getty)

The paper was later accepted for publication in Science Advances.

And while it still isn’t accurate enough to determine cases alone, the team believes it can help prioritise leads in tricky cases.

A history of fingerprinting

The use of fingerprints to identify people has been around for thousands of years.

Ancient Babylonians used fingerprints pressed in clay to record business transactions.

In China, fingerprints may have been used to establish identity as early as 300 BCE.

However, using fingerprints to crack criminal cases only began in the 19th Century.

But as is often the case with AI, the team wasn’t exactly sure how the system was determining which were pairs and which weren’t.

After carefully assessing the program’s decision process, the team thinks it is using a new kind of forensic marker.

‘The AI was not using “minutiae”, which are the branchings and endpoints in fingerprint ridges – the patterns used in traditional fingerprint comparison,’ said Mr Guo. ‘Instead, it was using something else, related to the angles and curvatures of the swirls and loops in the center of the fingerprint.’

Columbia engineering undergraduate Aniv Ray and PhD student Judah Goldfeder, who helped analyse the data, noted that their results are just the beginning.