AI can link two separate fingerprints from same person
Scientists say the ability to link the prints of individual fingers from the same person could help in reopening cold cases and overturning wrongful convictions.
Researchers have found that it is possible to detect when two different fingerprints come from the same person, shocking forensic scientists.
They say the new technology could help to resolve cold cases and overturn wrongful convictions.
It is a cornerstone of forensic science that all fingerprints are unique, even among the 10 fingers on an individual’s hands.
This means that if a burglar had left a fingerprint from their index finger at one crime scene and a print from their little finger at another, it was thought impossible to tell whether both came from the same person.
Academics from the United States now claim to have “shattered” this understanding. They performed a computer analysis of 60,000 fingerprints, feeding in pairs of prints into an artificially intelligent system called a “deep contrastive network”.
The researchers found their system learnt to detect with 88 per cent accuracy similarities between any two prints taken from different fingers belonging to an individual. They found that it “performed similarly across genders and races”.
They added that fingerprint comparisons usually focus on the “branchings and endpoints” in the ridges on a fingerprint but that the artificially intelligent system instead focuses on the “angles and curvatures of the swirls and loops” in the middle of a print.
The findings were deemed so surprising that their study was rejected by an unnamed forensic science journal, with researchers being told by one anonymous reviewer that: “It is well known that every fingerprint is unique.”
The researchers, from Columbia and Buffalo universities, then trained their system with more data and submitted it to a general science journal but again saw the paper rejected.
“I don’t normally argue editorial decisions, but this finding was too important to ignore,” said Hod Lipson, a professor at Columbia University and senior author of the study. “If this information tips the balance, then I imagine that cold cases could be revived and even that innocent people could be acquitted.”
It was ultimately accepted for publication by the Science Advances journal.
The first police fingerprint files were opened in Argentina in 1891 and the first criminal conviction took place a year later. Fingerprints were first used in police investigations in England and Wales in 1901 when the practice was formally accepted by the courts.
Asked how their AI system had spotted similarities that had eluded forensic scientists for 123 years, Gabe Guo from Columbia said: “The AI was not using ‘minutiae’, which are the branchings and endpoints in fingerprint ridges – the patterns used in traditional fingerprint comparison. Instead, it was using something else, related to the angles and curvatures of the swirls and loops in the centre of the fingerprint.”
Lipson said it showed how AI, when “given a fairly plain dataset that the research community has had lying around for years, can provide insights that have eluded experts for decades”.
Guo started the research while only a first-year student. Lipson said: “Even more exciting is the fact that an undergraduate student, with no background in forensics whatsoever, can use AI to successfully challenge a widely held belief of an entire field.”
They said that further research would be needed “before the technique can be used in practice” by police.
The Times
More Coverage
Is Trump a fascist? He’s too lazy to be a committed zealot
Two highly credible former aides have raised the rhetorical stakes about the dangers of re-electing the Republican.
One in four Americans fear civil war after poll
Fears that violence is very or somewhat likely to break out after the presidential election are shared by 27 per cent of American adults across the political divide.