Summary: Recent AI research conducted by undergrad researchers at Columbia Engineering challenges the long-held belief that all fingerprints are unique. Using a deep contrastive network and a US government database of 60,000 fingerprints, the researchers found that while the branching and endpoints of fingerprints may vary, the angles and curvature at the center can be the same across an individual. The AI system developed by the team was able to identify if prints belonged to the same person with an accuracy of 77 percent. The system’s accuracy increased when multiple pairs of prints were presented. The researchers discovered that the AI system was identifying the angles and starting points of the ridges, a previously unexpected forensic marker. This new information could potentially help prioritize leads in ambiguous cases.
In a surprising turn of events, recent AI research challenges the widely accepted belief that all fingerprints are unique. Undergrad researchers at Columbia Engineering undertook a study using a deep contrastive network and a US government database of 60,000 fingerprints to investigate the commonalities in fingerprints. Contrary to popular belief, the researchers found that while the branching and endpoints of fingerprints may vary, the angles and curvature at the center can be the same across an individual.
Using the deep contrastive network, the researchers fed pairs of fingerprints to a neural network, some belonging to the same person and others to different individuals. The AI system developed by the team was able to accurately identify if prints were from the same person with a 77 percent success rate. Interestingly, the system’s accuracy improved when multiple pairs of prints were presented.
The researchers initially struggled to understand how the AI system was able to identify the matching prints. To human eyes, the fingerprints appeared dissimilar. However, upon studying the decision process of the AI system, the researchers discovered that it was focusing on the angles and starting points of the ridges, providing an unexpected forensic marker for identifying fingerprints.
Engineering Professor Hod Lipson, recognizing the potential implications for cold cases, pushed to have the research published. While the accuracy of the AI system is not sufficient to officially determine a case, it is believed that it can assist in prioritizing leads in ambiguous situations.
Contrary to the notion that AI simply regurgitates existing knowledge, this research demonstrates how even a relatively simple AI system, given a plain dataset, can provide insights that have eluded experts for decades. The findings of this study will be published in Science Advances on January 12 at 19.00 UTC, offering an intriguing perspective on the uniqueness of fingerprints.
The source of the article is from the blog japan-pc.jp