A new artificial intelligence algorithm has found that the fingerprints of a single person may be more similar than previously believed. This discovery could have major implications for forensic investigations.
A research team at Columbia University developed an AI system that analyzed a database of 60,000 fingerprints. They found the AI could predict whether two prints came from the same person with 77% accuracy, even if they were from different fingers.
This challenges the assumption that prints from separate fingers have few similarities. If prints found across different crime scenes show unexpected commonalities, investigators may now be able to connect them to one perpetrator.
The AI focuses on the central swirls and loops of fingerprints rather than traditional comparison points like endpoints and bifurcations. Study leader Dr. Hod Lipson believes this fresh perspective allowed it to spot new connections.
“AI can find patterns that even domain experts miss,” said Dr. Lipson. “This could help crack cold cases or clear innocent suspects.”
While our own prints may be more alike than thought, the chance of matching another person remains infinitesimally small—estimated at 1 in 64 billion. So don’t worry about someone stealing your identity with a fake fingerprint just yet.
Factors like nutrition and blood flow in utero cause variation even between identical twins. Meanwhile, animals like gorillas and koalas have unique prints just like humans.
This new discovery opens up possibilities for re-examining fingerprint evidence, using AI’s new perspective to spot connections investigators previously missed.