GLOBAL
Rachel Lei of Heritage High School
31 March 2024
The interconnections of red thread mark separated pieces of evidence, each one cascading on the last. This evidence and how professionals perceive it rely on established understandings of the world.
One of these seemingly well-understood ideas is the nature of human fingerprints, each person having their own special set of curves on their hands. However, the trail of the thin, red thread has led students at Columbia University to discover that fingerprints might not be as unique as people previously thought.
Recently, advancements in fingerprinting emerged in automation and digital identification, where devices such as smartphones use sensors as a substitute for the traditional passcode.
According to a report published by the Federal Bureau of Investigation (FBI), fingerprint identification systems are “based primarily on the minutiae, or the location and direction of the ridge endings and bifurcations.”
In other words, fingerprint technology commonly studies and identifies the key patterns in skin, where presumably unique ridge patterns occur. In the criminal justice field, these technologies are essential for matching up individuals in crime cases, as well as protecting important information behind digital systems.
Thus, innovations in the field were historically focused on the identification of fingerprints and unique patterns, not on developing technology to find similarities. But following this trail of crumbs, Columbia Engineering student Gabe Guo developed a series of experiments to challenge seemingly established facts by utilizing emerging artificial intelligence (AI) technologies.
Through AI, a small percentage of fingerprints in a given dataset bore similar features to one another. Here, AI would occasionally conclude that fingerprints from different people were supposed to be fingerprints from the same person’s hands.
According to Holly Evarts from Columbia University who reported on the study, this was because the AI wasn’t analyzing fingers’ minutiae, but instead was looking at “the angles and curvatures of the swirls and loops in the center of the fingerprint.”
In forensic science, this technique would allow for greater accuracy in identifying suspects and comparing fingerprint evidence on-field. At the same time, however, the study acknowledges that the experiments’ results do provide several limitations.
The first limitation relates to the implications of “intra-person fingerprint similarity.” Because fingerprints could potentially be identified to more than one person, more caution will be needed if AI were to be used in the field. Researchers in the study advise of the development of alternative architectures to mitigate these results.
Another limitation explored in the paper is the relative ineffectiveness of the technology itself. It states that the AI system used is still “markedly below that of state-of-the-art systems designed for same-finger matching,” so the discoveries may not even apply to the field until the future. Because of its lower-level performance compared to current systems, the AI system used in the study would thus be inappropriate to use for authenticating evidence in court situations.
Despite these constraints, these findings remain significant in the fact that its increased accuracy in identifying fingerprint patterns would someday help “narrow down the candidate list” created by existing fingerprint identification systems. Not only would this be critical in the criminal justice field, but it would also advance password systems that rely on fingerprint data.
Due to technological nuances, this red thread may already have led to a dead end. But by continuing to roll out the thread, scientists could potentially pin it to a new piece of evidence on the corkboard.
The Brentwood Youth Press 