AI recognizes the fingerprints of the same person: applications in the crime world

Until now it has been believed that each fingerprint would be unique and different from the other: not only from individual to individual, but also between the different fingers of a single person. At least that’s what we thought so far. But science moves forward, and according to one study based on artificial intelligence born from a team led by Gabe Guoresearcher at Columbia University, it would seem that there is a strong one similarity between the prints of different fingers of the same individual.

The discovery, published in the prestigious magazine Science Advancesit might have a significant impact on future forensic investigations. Until now, in fact, if a murderer had left different finger prints on two separate crime scenes, the police would have focused on the presence of two criminals (precisely because it has always been thought that each print was “unique”), but this new technology will allow to understand that it would be a single person, thus significantly reducing the number of suspects.


What emerges from the analysis of fingerprints with AI

The team of researchers at Columbia has trained artificial intelligence using a databases of 60,000 fingerprints, examined one by one by an artificial intelligence system called deep contrastive network. The prints were inserted in pairs: both belonging to a single person and to two different people. With time, artificial intelligence was able to identify fingerprints belonging to the same individual with an accuracy of 77%. It might still seem like a low percentage, but the precision is already 10 times higher than that of current forensic investigations. According to the Columbia team, the discovery they just made could potentially increase forensic efficiency by more than 10 times if the AI ​​had other fingerprints from the same hand at its disposal.

But from a practical point of view, How did the AI ​​understand that two apparently so unique and different fingerprints belong to the same person? Rather than focusing on the “minutiae” – the branches and endpoints in the ridges – she focused on the angles and curvatures of the whorls and rings in the central part of the footprint.


Could there be two people with identical fingerprints?

It should be pointed out that even if our finger prints are not that different from each other, it is highly unlikely that there is anyone out there with fingerprints matching ours: it is estimated that the probability that two people share the same fingerprints is less than 1 in 64 billion. Even homozygous (therefore identical) twins have different footprints: this is because the ridges in the footprints are not only the result of our DNA, but also of the development of the fetus in the womb, blood pressure and the growth speed of the fingers.


The applications of AI in the world of crime

The team of researchers has already applied this new method by simulating a law enforcement investigation, and found a reduction in the number of clues by more than an order of magnitude. Aniv Ray, an engineer from Columbia, referring to the application of the discovery in the forensic field, stated:

We are now working to expand the database and obtain even more precise and accurate results. Imagine how well this AI will work once it works on a database with millions of fingerprints.

The most interesting aspect is certainly the possibility of exonerate suspected but innocent people or even to create new leads for unsolved cases: by expanding the number of fingerprints present in the database, in fact, the probability of finding the true perpetrator of a crime increases.