By 
In a groundbreaking development, researchers at Columbia University utilized an AI algorithm to scrutinize fingerprints, challenging the conventional belief in their absolute uniqueness. The findings suggest that certain patterns on fingerprints are not as distinct as previously thought.
The research team, comprised partly of individuals without prior experience in forensic biometrics, employed a deep contrastive network for this unconventional analysis. Led by an undergraduate who initiated the work as a first-year student, the researchers directed the algorithm towards a NIST database containing 60,000 fingerprints for intra-hand matching.
The results indicated an accuracy of 77 percent for a single pair of prints and peaked at 88 percent for multiple prints. While these figures might not currently meet the standards for courtroom use, the research could guide investigators in prioritizing leads, yielding unexpected outcomes. The university emphasizes that training the system on millions of fingerprints could significantly enhance its accuracy.
Despite the potential implications of the research, a “well-established” forensics journal initially rejected the paper, citing the established uniqueness of every fingerprint. However, the paper has now been accepted by the esteemed journal Science Advances.
The novel approach taken by the biometric algorithm focused on the curves and angles of swirls at the center of the fingerprint, commonly known as “the singularity.” This method deviates from the traditional minutiae-based approach, which examines the branching and end points of print ridges. Strikingly, the results demonstrated consistency across subjects of various races and genders.
The researchers’ next steps involve analyzing data from a more diverse sample of individuals, opening new avenues for understanding and refining the capabilities of forensic biometrics.
Comments