Engineers from the American universities of Columbia and Buffalo have created a new fingerprint analysis using artificial intelligence (AI) that overturns the long-held belief in forensic medicine that no two fingerprints are the same, not even on different fingers of the same person.
The finding, reported this Wednesday by the journal Science Advance, has shown, with a reliability of 99.99%, that the fingerprints of any two fingers of the same person are much more similar than previously thought.
Fingerprints are essential in crime laboratories to solve cases, and in billions of mobile phones around the world for digital authentication, although, for now, all technologies in this field are designed under the premise that they do not There are two identical fingerprints.
To date, fingerprints are not useful in situations where the available prints are from fingers other than those recorded, such as at a crime scene. However, a study promoted by Gabe Guo, an engineering student at Columbia, along with other researchers from the same university and the University of Buffalo, has shown that it is possible to overcome this limitation by analyzing characteristics of fingerprints that have not been studied until now.
Guo and his colleagues found a public US government database with about 60,000 fingerprints and entered them in pairs into an artificial intelligence-based system known as a deep contrast network. Sometimes the pairs belonged to the same person (but with different fingers) and sometimes to different people.
Engineers, without prior forensic knowledge, extracted fingerprint representation vectors from 525,000 images using deep neural networks and made a surprising discovery: fingerprints from different fingers of the same person are extremely similar. They discovered that the orientation of the ridges near the center of the prints explained much of this similarity, and that this pattern holds for all pairs of fingers from the same person.
The model has been successfully tested with women of different genders and racial groups.
“We hope that this additional information can help prioritize leads when there are many possibilities, exonerate innocent suspects, or even create new leads for unsolved cases,” Guo said in a statement from Columbia University. The researcher also emphasizes that his discovery could improve the convenience and accessibility of digital authentication techniques.