AI Helps Challenge Long-Held Beliefs About The Uniqueness of Fingerprints

In a groundbreaking study that could redefine forensic science as we know it, a team of engineers from Columbia University has developed an artificial intelligence system that challenges the long-held belief in the uniqueness of fingerprints from different fingers of the same person.

For years, crime shows like “CSI” and “Midsummer Murders” not to mention real-life criminal investigations, have leaned heavily on fingerprint analysis as definitive evidence. Finger prints are well known to be unique from person to person providing a simple and robust method for identifying suspects. But, problematically, they were also thought to be unique from finger the finger. So if a suspect leaves prints from different fingers in two different crime scenes, these scenes are very difficult to link, and the cases can go cold.

However, in new research researcher published today, led by Columbia Engineering undergraduate Gabe Guo, it is suggested that we’ve been analysing fingerprints all wrong.

Guo, a senior with no prior background in forensics, stumbled upon a public U.S. government database containing about 60,000 fingerprints. Utilising a deep contrastive network—a sophisticated type of AI—he fed pairs of fingerprints into the system. Some pairs were from different fingers of the same person, while others were from different individuals.

Saliency map highlights areas that contribute to the similarity between the two fingerprints from the same person. Gabe Guo, Columbia Engineering

The findings are nothing short of astonishing. Initially, the AI system’s accuracy in matching individual prints from the same person was 77% (which in of it’s self is much higher than previously understood). However, when multiple pairs were analyzed, the accuracy soared, suggesting a potential tenfold increase in forensic efficiency.

A troubling publication

The path to publication was fraught with skepticism and disbelief. The forensics community, entrenched in the idea of fingerprint uniqueness, initially rejected the findings. An anonymous expert reviewer dismissed the possibility of detecting similarities in fingerprints from the same person. But the team, undeterred, refined their AI model, feeding it more data and improving its accuracy. And inspired by their high quality data and study by Professor Lipson (James and Sally Scapa Professor of Innovation) appealed the original decision.

“I don’t normally argue editorial decisions, but this finding was too important to ignore. If this information tips the balance, then I imagine that cold cases could be revived, and even that innocent people could be acquitted.” 

Professor Lipson

After more back and forth, the paper was finally accepted for publication by Science Advances.

A new kind of marker

The AI, it turns out, was identifying a new type of forensic marker in fingerprints, focusing not on the traditional minutiae but rather on the angles and curvatures of patterns within the prints.

The implications of this discovery are profound. It could lead to the reopening of cold cases and, more importantly, the exoneration of wrongly accused individuals. While the current accuracy level of the AI system is not yet sufficient to make official forensic determinations, it can significantly aid in prioritizing leads in ambiguous cases.

This study also highlights the need for more inclusive and broader datasets in AI research to avoid potential biases, a point the team acknowledges and plans to address in future work.

Perhaps most strikingly, this research exemplifies the transformative potential of AI in established scientific fields. As Professor Lipson notes, the success of an undergraduate student in using AI to challenge a widely held belief underscores the imminent wave of AI-led scientific discoveries by non-experts. This democratization of discovery, propelled by AI, is set to revolutionize not just forensics but potentially every field of scientific inquiry.

The Columbia team’s work is a testament to the power of interdisciplinary collaboration and the transformative potential of AI. It marks the beginning of a new era in forensic science, one where AI could play a pivotal role in unraveling the mysteries that have long eluded human experts.

You can read Columbia University’s full paper titled “Unveiling Intra-Person Fingerprint Similarity via Deep Contrastive Learning” in Science Advances here. Their work is part of a joint University of Washington, Columbia and Harvard NSF AI Institute for Dynamical Systems, aimed to accelerate scientific discovery using AI, a project we eagerly anticipate hearing more from.

Staff Writer

Our in-house science writing team has prepared this content specifically for Lab Horizons

Leave a Reply

Your email address will not be published. Required fields are marked *