Compiled by | Xu Rui
The unique patterns on human fingertips form fingerprints. Currently, no two individuals have been found to have the same fingerprints, so fingerprints are unique identifiers for each individual and are widely used in fields such as identity recognition and criminal investigation.
On January 12th, a study published in Science Advances used artificial intelligence (AI) to find similarities between different fingerprints of the same individual, thus enabling the accurate identification of whether fingerprints left by different fingers belong to the same person. This aids criminal investigators and forensic experts in determining whether an individual has appeared at different crime scenes.
The current technology can only match fingerprints left by the same finger. However, previous research has suggested that there may be fundamental similarities between all fingerprints on an individual's fingertips.
In this study, Gabe Guo and colleagues from Columbia University in the United States extracted over 50,000 fingerprints from a public database of nearly 1,000 individuals, obtained from the National Institute of Standards and Technology and the State University of New York at Buffalo. These fingerprints belonged to deceased individuals or individuals without identity information. They used these fingerprints for training, allowing the machine learning model to identify whether fingerprints from different fingers belong to the same person.
After completing the training, the research team tested and evaluated the model on a collection of over 7,000 fingerprints from approximately 150 individuals, with accuracy scores ranging from 0 to 1. The researchers found that scores above 0.75 indicated that the model could reliably identify whether fingerprints from different fingers belong to the same person.
"This technology can improve the efficiency of forensic investigations," said Ralph Ristenbatt from Pennsylvania State University. "Sometimes, fingerprints found at multiple crime scenes may not match anyone in the database. By utilizing the above model, it may be possible to determine if fingerprints left at different specific crime scenes belong to the same individual."
However, Guo stated that the current model's accuracy is still not high enough to assist in court judgments.
"If we want to use this model for legal purposes, it needs to be trained on a larger database," said Hod Lipson from Columbia University, a member of the research team.
Related paper information:
https://doi.org/10.1126/sciadv.adi0329
China Science Daily (January 17, 2024, 2nd Edition, International)Editor | Zhao LuLayout | Guo GangPlease scan the QR code below for identification (3 seconds)
