Human fingerprints turned out not to be unique: a revolutionary discovery by scientists
The discovery of the fact that fingerprints are unique to each individual person has led to a real revolution in forensics. Scientists have also confirmed this fact - every human body is unique. Moreover, they suggested that among the more than 100 billion people who have ever lived on Earth, fingerprints have never been repeated. It was believed that the pattern on each individual finger was not repeated by any other. So it was believed that each palm had 5 completely different prints.
But recently, scientists from Columbia University in New York have called this axiom into question. According to Science Alert, a study led by senior engineer Gabriel Guo showed that the fingers of one person can have almost identical prints.
The researchers used artificial intelligence for their work. They asked it to examine pairs of prints. In some cases, they belonged to the same person, in some - to different ones. The neural network's task was to determine whether it was possible to detect similarities in patterns on different fingers of the same person. And the conclusion was revolutionary: with a very high degree of probability, such prints are, if not identical, then as similar as possible. At the same time, the prints of different people are indeed radically different from each other.
For the study, the researchers used a publicly available US database containing about 60,000 fingerprints. It consists of sets of 10 fingers for each individual. Guo and his colleagues decided to try to see if artificial intelligence could detect the similarity of prints belonging to the same person. To do this, they used a dual neural network and instructed the machine to examine the image not according to a typical pattern, but to consider the angles and curvature of the curls and loops in the center of the fingerprint.
Over time, it turned out that the neural network was getting better at distinguishing pairs of fingerprints belonging to the same person from pairs that contained fingerprints from different people. It turned out that on one palm and even a pair of palms, if they belong to the same person, the pattern has enough similarities to establish a match. In particular, the orientation of the stripes in the center of the print was the same for different fingers of the same person.
The researchers admit that their discovery cannot yet be used in practice. However, they expressed confidence that further training of the neural network will allow it to accurately identify the fingerprints of different people based on this similarity. And this could lead to another breakthrough in forensics, significantly increasing the accuracy of fingerprinting. In particular, it will be possible to review court verdicts that raise doubts and rehabilitate people who have been unjustly accused.
Among the possible applications of the approach is the possibility of using any finger to unlock a mobile phone equipped with a fingerprint scanner. But now it is necessary to conduct deeper training of artificial intelligence to significantly improve the accuracy of its conclusions.