Fingerprints, artificial intelligence reveals the similarities between the fingers of the same person – WWN

Fingerprints, artificial intelligence reveals the similarities between the fingers of the same person – WWN


Of Massimo Sideri

Thanks to a database of 60 thousand fingerprints, the system was able to understand when two fingers belonged to the same subject, with a probability of 77% to 99%.

The uniqueness of Fingerprints , in itself, is an axiom: it has never been possible to prove the opposite, so we accept it as given. In fact, for now the empirical evidence has demonstrated it, as they say in American trials, beyond any reasonable doubt. Even in monozygotic twinswhich are born from the same egg and therefore have the same DNA, fingerprints are different, because they are influenced by environmental conditions and epigenetics. Now this axiom has been questioned for the first time and the culprit could only be artificial intelligence. Let’s start by saying that everything is yet to be confirmed. We are at the beginning of a research that has just been published on Science Advances, only after a very complex editorial life precisely because it is a scientific result taken for granted. Technically speaking we are faced with clues, not proof.

Similarity between fingers

For now Agatha Christie and a century of detective stories are actually safe: no one questioned the uniqueness of the footprints between different people. But in the study, born by a team led by Gabe Guo, an engineering researcher at Columbia University who had no forensic experience before obtaining this clue, it is hypothesized that there may be a similarity between the prints of different fingers of the same individual. Imagine a killer committing a murder leaving a right thumb print. Now imagine the same serial killer committing another murder at another time leaving a left index finger print this time. Until now, forensic science took it for certain that there could be no relationship (which is why we see in movies that when you take fingerprints with ink it creates a ten finger card).

Uniqueness of the footprints

Guo instead trained an AI with a database of 60 thousand fingerprints finding that the machine was able to identify when two fingers belonged to the same subject with an increasing probability (from 77% to 99% in some cases presented in the study). Not sure enough. But this is where the request for publication comes from be able to continue with training through larger databases. How is it possible that there is a difference between traditional forensic methods and AI? To understand this, it is worth diving for a moment into the fascinating history of fingerprints. The ridges we all have on our fingertips they were first described in 1665 by an Italian, Marcello Malpighi, the father of microscopic observation in anatomy. But we have to get to an article published on Nature in 1880 signed by Henry Faulds, doctor and missionary in Japan, to read about uniqueness of fingerprints and their possible use in crime scenes to identify the culprit.

New analysis methods

However, it has only recently been discovered that the mechanism of this uniqueness is very close to the one that in nature allows zebras to have a coat with streaks that are never the same between two different animals. A hypothesis that had been anticipated by Alan Turing, the prolific father of artificial intelligence in 1950. For decades, forensic science has been concerned with demonstrating the uniqueness of human footprints by studying the so-called minutiae, that is, the terminal part of the lines that create small valleys on our fingers. According to Guo, the trained AI instead changed its methodology, studying the vortices more in the central part of the footprints. For the enthusiastic researcher from Columbia University it would even be proof that AI can find new methods of scientific analysis. But a little early for that. For now it remains a very powerful tool at the service of science.

Corriere della Sera also on Whatsapp. sufficient click here to subscribe to the channel and always be updated.

January 13, 2024 (modified January 13, 2024 | 2.44 pm)



Source link