The prints from two fingers on the same hand may appear different, but AI can spot fundamental similarities
Andrey_Kuzmin/Shutterstock
Artificial intelligence can accurately identify whether or not fingerprints left by different fingers came from the same person. This could help forensic investigators uncover if one individual was present at separate crime scenes.
Current technologies are only capable of matching fingerprints left by the same finger. But previous studies have hinted there may be fundamental similarities between all of a person’s fingertips.
So, Gabe Guo at Columbia University in New York and his colleagues trained a machine learning model to determine if it could identify whether fingerprints from different fingers belong to the same person. In the training, they used more than 50,000 fingerprints from almost 1000 people. Samples came from public databases at the US National Institute of Standards and Technology and the University at Buffalo in New York. All fingerprints belonged to deceased individuals or had been de-identified from people still alive.
The team then tested the trained model on a separate set of more than 7000 fingerprints from approximately 150 people. They evaluated the model using a statistical measure that estimates accuracy on a scale of 0 to 1. The researchers found their model scored greater than a 0.75, suggesting it reliably identifies whether or not prints from different fingers belong to the same person.
This technology could enhance the efficiency of forensic investigations. “It could be useful in cases where you have fingerprints found at multiple crime scenes, which don’t match to anybody in databases,” says Ralph Ristenbatt at Pennsylvania State University. “Could the person who left the fingerprints at this particular crime scene be the same person who left [different] prints at this other crime scene?”
However, “right now the accuracy isn’t high enough [for this model] to decide this in the court of law”, says Guo.
“If this thing is used in practice, for legal purposes, it should be retrained with a [bigger] database,” says Hod Lipson, also at Columbia University and part of the research team.
Topics:
