A new study from Columbia University, published in the journal Science Advances, challenges the long-held assumption that human fingerprints are unique to each person. The research relied on artificial intelligence to uncover similarities between fingerprints, even those belonging to the same person.
Lead author Gabe Guo, an engineering student at Columbia, fed a database of 60,000 fingerprint pairs into a deep learning algorithm. The AI system was eventually able to predict with 77% accuracy whether two prints came from the same person or not. It did this by analyzing the patterns and curves in the central part of fingerprints.
While the technology is not yet accurate enough to be used in court, the authors argue it could help generate leads in cold cases by matching crime scene prints to different fingers on file for suspects. However, some forensics experts criticize the study, saying it has not truly disproven fingerprint individuality and that print similarity within persons was already known.
Guo stands by the research, claiming it goes further than prior work and hoping it sparks additional discoveries using AI to reexamine things “hiding in plain sight.” The team has open-sourced the model’s code for public scrutiny. The debate continues around what this means for the uniqueness of human fingerprints in identification.