- AI shows that each person’s fingerprint is not unique
- The breakthrough allowed thousands of cold cases to be solved
Thousands of cold cases could be solved thanks to a breakthrough in fingerprint analysis through artificial intelligence.
A computer using an artificial intelligence system has shattered decades of wisdom that each person’s fingerprint is unique.
So if a criminal left a fingerprint at one crime scene and a print of his index finger at another, there would be no way to connect the two.
The breakthrough came when a Columbia University student tried to see if artificial intelligence could find connections between apparently very different fingerprints of the same person.
To test the idea, Gabe Guo, an engineering graduate with no background in forensics, presented a computer with images of about 60,000 fingerprints in pairs.
Thousands of cold cases could be solved thanks to a breakthrough in fingerprint analysis through artificial intelligence. A computer using an artificial intelligence system has overturned decades of wisdom that each person’s fingerprint is unique (stock image)
In some cases the fingerprints are said to come from two different fingers on one person’s hand, and in other cases from different people.
Over time, the computer managed to discover giveaway patterns, which meant that two fingerprints that looked very different came from the same hand – something that had never been discovered before.
Writing in Science Advances, Guo and colleagues write: ‘Our key discovery is that fingerprints from different fingers of the same person show strong similarities; these results apply to all finger combinations, even to different hands of the same person.’
The findings were initially dismissed by the forensic community.
A well-known forensic journal dismissing the study, with an anonymous expert reviewer and editor concluding that ‘it is well known that each fingerprint is unique’ and therefore it would not be possible to detect similarities even if the fingerprints came from the same person .’
But Mr. Guo and his colleagues persisted.
Hod Lipson, professor of engineering at Columbia, said: ‘I don’t normally argue editorial decisions, but this finding was too important to ignore.
“If this information tips the balance, I imagine cold cases could be revived and even innocent people could be acquitted.”
The salience map highlights areas that contribute to the similarity between the two fingerprints of the same person
One of the sticking points that led to the rejection of the research finding is that it was not clear what information the AI used to connect seemingly unrelated finger pins, which had eluded decades of forensic analysis.
The team concluded that the AI had identified new patterns in the edges of the fingerprint’s centers that had not been seen before.
Professor Lipson said the research was an example of a new insight from AI.
He said: ‘Many people think AI can’t really make new discoveries – that it just regurgitates knowledge.
‘But this study is an example of how even a fairly simple AI, given a fairly simple dataset that the research community has had for years, can yield insights that have eluded experts for decades.
He added: ‘Even more exciting is the fact that a student, without any background in forensics, can use AI to successfully challenge a widely held belief in an entire field.
“We are about to experience an explosion of AI-led scientific discoveries by non-experts, and the expert community, including academia, must prepare.”
The discovery has another potential use: for computer fingerprint identification. Some computer laptops and security systems identify people using a fingerprint. But if the finger used to make the original print is damaged (for example, because it is bandaged), it means the user is locked out.
The new system allows another finger to be used instead.