In many crime-related TV shows or films, fingerprints are often considered crucial for establishing the relationship between the perpetrator and a crime. It’s an accepted idea that if someone leaves fingerprints from different fingers at different crime scenes, it would be difficult to get a match as fingerprints are believed to be unique. Now, a new study uses artificial intelligence (AI) to show that this might not be the case and they might be similar.
For the study, conducted by researchers from Columbia University School of Engineering and Applied Science, a government database of around 60,000 fingerprints was fed as pairs into an AI-based system known as a deep contrastive network, a press statement explained.
Some of the pairs belonged to the same person, but different fingers, and some were of different people. The AI system showed that fingerprints from different fingers of the same person have similarities and that researchers have been comparing fingerprints the wrong way.
The AI system, which was designed by modifying a state-of-the-art framework, reached an accuracy of 77% and sometimes higher when determining whether a single pair of fingerprints belonged to the same person or not. The researchers also realised that the AI was using a new kind of forensic marker.
“The AI was not using ‘minutiae,’ which are the branchings and endpoints in fingerprint ridges – the patterns used in traditional fingerprint comparison,” lead author Gabe Guo said in the statement. “Instead, it was using something else, related to the angles and curvatures of the swirls and loops in the centre of the fingerprint.”
According to the researchers, this discovery is an example of the huge potential of AI. While people think that AI cannot really make new discoveries and just regurgitates knowledge, this research shows how it can provide insights into things that experts haven’t figured out for decades, the researchers elaborated. The results were published in the journal Science Advances.
It’s also exciting that an undergraduate student, Guo, who has no background in forensics, could use AI to challenge a widely held belief of an entire field, the press statement added.