Neural network has learned to fake fingerprints to break into biometric systems

Scientists from the Polytechnic Institute of New York University have taught the neural network to forge fingerprints for further hacking of biometrics systems. About this writes The Guardian.

Biometric authentication systems do not scan a person’s entire finger, but only a part of it, and then compare it with stored user data. This means that fraudsters just need to recreate not even the entire human finger, but only part of the print.

In addition, in humans, many of the reliefs on the imprint are repeated, so an artificially made imprint containing several of the most common reliefs will definitely fit directly to several fingerprints.

The Neural Network DeepMasterPrints studied the database with a huge number of fingerprints and learned how to create such artificial fingerprints that correspond to several real ones at once. Artificial prints of all can fake up to 23% in a database with a permissible error of 0.1%. With an error of 1%, this figure increases to 77% of objects.

hightech.fm