AI Analyzing Fingerprints

They say that every fingerprint is unique, even from one finger to another on the same hand. But what impact would it have on the criminal justice system if that wasn’t quite the case. We’ll explore advances in forensics in the U.S. National Science Foundation’s “Discovery Files”.
Fingerprints are the gold standard for identification in forensic science and have become the default authentication used to access many of today’s smart phones and laptops.
The assumption has long been that no two fingerprints are alike, even those from different fingers of the same individual. But a team including NSF-supported researchers at Columbia University have found some similarities.
Using a publicly available database of fingerprints and advanced artificial intelligence neural networks, they were able to train a system to identify similarities across all pairs of fingers for an individual.
They found that the system had developed a new kind of forensic marker, bypassing the branching and endpoints used in traditional fingerprint comparison focusing instead on the angles and curvatures of the loops and swirls found at the center of a fingerprint.
While further development will be necessary before it’s ready for use by the legal system, this research could revolutionize forensic work, with the ability to potentially connect individual fingerprints collected from different crime scenes.
To hear more science and engineering news, including the researchers making it, subscribe to "NSF's Discovery Files" podcast.