AI discovers that not every fingerprint in unique, how?
It is widely acknowledged in the forensics community that fingerprints from different fingers of the same person–“intra-person fingerprints”–are distinct and thus unmatchable. A team lead by Columbia Engineering undergraduate senior Gabe Guo questioned this generally known assumption.
Guo, who has no prior experience with forensics, discovered a public US government database of over 60,000 fingerprints and fed the resulting pairs into an artificial intelligence-based system called a deep contrastive network. The pairings could belong to the same individual (but with different fingers) or to separate people.
As time passed, the AI system, that the team created by tweaking a cutting-edge framework, improved its ability to distinguish between supposedly identical fingerprints belonging to the same person and those that did not. When numerous pairs were shown, accuracy improved dramatically, potentially enhancing current forensic effectiveness by more than tenfold.
The project, a partnership between Hod Lipson’s Creative Machines lab at Columbia Engineering and Wenyao Xu’s Embedded Sensors and Computing lab at the University of Buffalo (SUNY), was published in Science Advances.
A New Forensic Marker To Capture Fingerprints
One of the sticky points was the following: What alternative information did the AI use to elude decades of forensic analysis? After carefully visualising the AI system’s decision process, the researchers concluded that the AI was employing a novel type of forensic marker.
According to Guo, a freshman at Columbia Engineering who started the study in 2021, the AI did not use ‘minutiae,’ which are the branchings and terminals in fingerprint ridges – the patterns employed in traditional fingerprint comparison. He claimed that it was instead using something else, namely the angles and curvatures of the swirls and loops in the centre of the fingerprint.
The team understands that the data may be biassed. The authors show evidence that the AI performs comparably across genders and races when samples are available. They do, however, point out that if this technique is to be implemented in reality, more thorough validation with datasets with larger coverage is required.
It is true that an undergraduate student with no prior forensics experience has been successful in contesting a commonly held assumption in an entire field using AI. We are going to witness an explosion of AI-driven scientific discovery by non-experts, and the expert society including academics, must prepare.”
Fingerprints and Crime Scenes
If, for example, an unidentified thumb print is discovered at crime scene A and an unidentified index finger print at crime scene B, the two cannot currently be forensically linked to the same person – but the AI tool may be able to detect this.
The Columbia University crew, none of whom had forensic credentials, admitted that additional investigation was required.
There were concerns about whether the markers the AI tool was focused on stayed the same based on how the skin twisted as it made contact with the print surface, as well as if they remained consistent over the course of a lifetime, as traditional markers do. However, this could be difficult to respond because the researchers are not certain of what the AI is doing, as is the situation with many AI-powered tools.
Fingerprints are formed before to birth. According to a study released last year, the genetic process underlying them may be similar to how animals such as zebras and leopards acquire their markings, an idea initially presented by codebreaker Alan Turing in the 1950s.
Conclusion
The conventional thinking in forensic science has long been that fingerprints from different fingers of the same person are distinct and cannot be matched. This remarkable study, on the other hand, significantly challenges that idea. The team is also aware of potential biases in their data and recognises the importance of additional validation across varied demographics for practical use.
This study demonstrates the enormous potential of AI in the field of science. Lipson observes that AI, even in its most basic versions and with readily available datasets, can reveal discoveries that have eluded specialists for decades.
To summarise, as AI evolves, it creates a new frontier in scientific research, allowing non-experts to contribute to important discoveries.This case signals an impending increase in AI-driven scientific achievements, calling the scientific community and academia to get ready for the next phase of innovation.
