Artificial Intelligence

Not every fingerprint is unique, AI discovers

11th January 2024
Paige West
0

Columbia University engineers have developed a new AI that contradicts a longstanding assumption in forensics: fingerprints from different fingers of the same person are not as unique as previously thought.

Contrary to popular belief and depictions in shows like ‘Law and Order’ and ‘CSI’, this discovery reveals that fingerprints from various fingers can be similar, suggesting we've been analysing them incorrectly.

Traditionally, forensics has considered ‘intra-person fingerprints’ (fingerprints from different fingers of the same individual) to be unique and unmatchable. However, a team led by Columbia Engineering undergraduate senior Gabe Guo, with no background in forensics, challenged this notion. Guo used a public US government database of around 60,000 fingerprints, feeding pairs into a deep contrastive network AI system. The pairs sometimes belonged to the same person (but different fingers) or different people.

The AI system, adapted from an advanced framework, improved its ability to discern when seemingly unique fingerprints were from the same person. The accuracy for a single pair reached 77%, and with multiple pairs, it significantly increased, potentially enhancing forensic efficiency by more than ten times. This project was a collaboration between Hod Lipson’s Creative Machines lab at Columbia Engineering and Wenyao Xu’s Embedded Sensors and Computing lab at the University at Buffalo, SUNY, and was published in Science Advances.

Upon submitting their findings to a forensics journal, the team faced rejection due to entrenched beliefs in the field. However, they persisted, improving the AI system with more data. Despite initial scepticism from the forensics community, the team chose to submit their manuscript to a broader audience. After a series of rejections and appeals, the paper was accepted by Science Advances.

The researchers identified a new type of forensic marker used by the AI, different from the traditional 'minutiae' (branchings and endpoints in fingerprint ridges). Instead, the AI focused on angles and curvatures in the centre of fingerprints.

Columbia Engineering senior Aniv Ray and PhD student Judah Goldfeder, who contributed to the data analysis, believe this is just the beginning. The team recognises the need for broader datasets to validate the AI's performance across different genders and races.

This discovery exemplifies the transformative potential of AI in established fields. Hod Lipson, co-director of the Makerspace Facility and James and Sally Scapa Professor of Innovation, highlights the significance of this research. He notes that AI can provide new insights from existing data and emphasises the remarkable achievement of an undergraduate student challenging long-held beliefs in a field using AI. This breakthrough suggests a forthcoming surge in AI-led discoveries by non-experts, reshaping the landscape of scientific inquiry and expertise.

Featured products

Upcoming Events

View all events
Newsletter
Latest global electronics news
© Copyright 2024 Electronic Specifier