Content
Page 11
Automating Detective Work Artificial intelligence is improving biometrics for the matching of fingerprints and facial recognition.

E VERY FINGERPRINT is be- lieved to be unique, mak ing it possible to identify an individual by matching a new fingerprint with an im- age on file, whether to unlock a mo- bile phone, access a bank account, or solve a murder. Fingerprint examin- ers, however, do not always agree on whether two fingerprint images match and, asked to recheck their work after several months, they sometimes do not even agree with themselves. That is leading to increased use of neural networks, powerhouses for identifying and matching patterns of all sorts, to automate and improve decisions about whether two fingerprints come from the same person.

A group of computer scientists de- cided to use neural networks to test the assumption that no two finger prints are the same. Using twin neu- ral networks, researchers from Co- lumbia University, Tufts University, and the State University of New York (SUNY) at Buffalo looked for similari ties between different fingerprints in a database from the National In- stitute of Standards and Technology (NIST). They applied deep contrastive learning, a technique in which the neural network compares images to learn which attributes are the same and which differ. If the distance be- tween the network's statistical rep- resentations of two images is above a given threshold value, the images are deemed to be different. Below that val- ue, they are ruled to be the same.