A longstanding belief in the unique nature of fingerprints is facing a challenge, according to groundbreaking research conducted by a team at Columbia University. Contrary to the widely held notion that each person's fingerprints are entirely distinct, an artificial intelligence (AI) tool developed at the U.S. university has demonstrated the ability to identify, with 75-90% accuracy, whether fingerprints from different fingers belong to the same individual.
The researchers, led by Prof Hod Lipson, a roboticist at Columbia University, trained the AI tool on a dataset of 60,000 fingerprints. Surprisingly, the technology seems to deviate from traditional forensic methods, focusing on the orientation of ridges in the center of a finger rather than the minutiae – the endpoints and forks of individual ridges. Prof Lipson admitted, "We don't know for sure how the AI does it," emphasizing the unconventional markers it appears to use, such as the curvature and angle of the swirls in the fingerprint's center.
While the results of the study have potential implications for biometrics and forensic science, the researchers caution that more investigation is needed. Graham Williams, a professor of forensic science at Hull University, noted that the assumption of fingerprint uniqueness has never been definitive, stating, "We don't actually know that fingerprints are unique." The AI tool could potentially bridge fingerprints found at different crime scenes, leading to new possibilities in forensic investigations.
However, the Columbia University team, lacking a forensic background, acknowledges the need for further research. The AI tool, while promising, is not currently deemed suitable for deciding evidence in court cases but may serve as a valuable tool for generating leads in forensic investigations.
Dr Sarah Fieldhouse, an associate professor of forensic science at Stafford University, expressed skepticism about the study's immediate impact on criminal casework. Questions persist about the stability of the AI tool's identified markers, especially regarding changes in skin contact with print surfaces and over a person's lifetime.
The study, which has undergone peer review, is set to be published in the journal Science Advances on Friday. The AI's potential to challenge the conventional understanding of fingerprint uniqueness adds a new dimension to discussions around biometrics and forensic science, even as uncertainties about the technology's mechanisms remain.