An algorithm capable of creating realistic enough fingerprints to fool electronic readers has been developed by researchers at New York University.
The fingerprints generated by the artificial intelligence system are so well designed that they can fool an electronic reader once in five, even if they do not correspond to real fingers.
To achieve this result, researchers assembled a database containing real fingerprints. The latter triggered a false positive only once in a thousand on average.
They then induced a network of artificial neurons to recognize the similarities between the different imprints and to create new ones by assembling the most common parts.
Scientists were looking to exploit a weakness of fingerprint readers. Due to the shape of the human fingers, these devices usually have to perform their analysis from a fingerprint fraction. They then compare this fingerprint to the images stored in their database. This results in a higher error rate than if a complete image had been used.
By assembling the most common pieces of fingerprints, the researchers managed to trigger these false positives with enough consistency to seriously question the safety of places and devices protected by fingerprint scanners.
False fingerprints generated by the algorithm, however, are more effective on drives connected to large databases, such as those protecting the entrance doors of buildings. This is because synthetic fingerprints are more likely to match an actual footprint if there are hundreds or thousands in the system than if there is only one, such as on your phone .
Tricia Nelson is a skilled writer at 3Zebras, known for her insightful articles on technology, science, and gaming. She covers a wide range of topics, from the potential of nanotechnology in mimicking cobwebs to exploring new discoveries in astronomy. Tricia’s work stands out for its clear explanations and engaging style, making complex subjects accessible to all readers. Her contributions have made her a respected voice in the 3Zebras community,