Through the use of artificial intelligence, the uniqueness of fingerprints may not be completely true. The reality might be a little more complex.
A recent study from Columbia University in Canada has challenged this firmly held belief in the field of forensic science. The latter brings together all the analysis methods used for investigative work, which the scientific police apply in particular in their investigations. No, our fingerprints would not be unique! A team of researchers managed to analyze 60,000 fingerprints using a tool powered by artificial intelligence. Result: some of them had rather surprising matches. Another area that AI could disrupt?
AI and fingerprints: a new vision
The study was carried out under the supervision of Professor Hod Lipson, a roboticist at Columbia. Their tool demonstrated 75 to 90% accuracy in associating prints from different fingers to the same individual. A fairly high success rate, which would suggest that each imprint is not really unique.
Unlike traditional methods, the AI focuses on the orientation of the ridges in the center of the fingers. “ It is obvious that it does not use the traditional markers used for decades in forensics » emphasizes Professor Lipson. On the other hand, he admits: “ we’re not entirely sure how AI achieves this “. A surprising discovery, which led Lipson and his team to verify several times whether the results were true.
Other experts seem a little less surprised by the news. This is the case of Graham Williams, professor of forensic science at the University of Hull. Indeed, according to him, there has never been absolute proof that fingerprints were specific to each individual. “ We don’t really know if the prints are unique (…) All we can say is that to our knowledge, no identical fingerprints have been observed in two different people “. All the nuance is there.
Implications and limitations of this discovery
The fallout from this discovery could strongly influence forensic science or biometrics. There may come a day when AI will be able to link unidentified prints at different crime scenes. It would be a small revolution for the field of criminal investigations!
However, the Columbia team lacks forensics experience. According to their point of view, more in-depth research is needed to confirm their discovery. Ideally, an AI tool should be trained on a larger database. Gabe Guo, an undergraduate student, explains: “ our tool is not reliable enough to be used as evidence in legal cases, but it is useful for generating leads in forensic investigations “.
Other researchers, such as Dr Sarah Fieldhouse (professor of forensic science), from Staffordshire University, remain more measured about this discovery. She believes that, in the immediate future, this study will not have any real legal impact for the moment. This will in all cases be published in the journal Science Advances on January 13 and will still open up new perspectives on the analysis of fingerprints.
Source : BBC
6