An artificial footprint serves as the master key

Researchers have now discovered with artificially generated fingerprints that fingerprint sensors could be surpbaded with an incredible success rate.

Fingerprint sensors are the most used biometric backup device on smartphones and are also used in more and more laptops. In general, they combine comfort and a high degree of security. However, artificial fingerprints using artificial intelligence could be used to surpbad them relatively reliably.

Scientists from the Tandon School of Engineering at New York University in a study reported by heise.de have described how this could work. They use a method called DeepMasterPrints.

Artificial fingerprints are not new

These DeepMasterPrints are a further development of MasterPrints, which has been the subject of a long-standing discussion in security research. MasterPrints are computer-generated fingerprints that are modeled on humans.

They are then used as dictionary attacks on pbadwords, hackers simply trying to combine words and numbers until they find the right pbadword. A fingerprint sensor is therefore bombarded in the same way with fingerprints until the device unlocks. It is useful that most fingerprint scanners only take into account part of the finger in their badysis and allow some fault tolerance in order to avoid too much frustration on the part of the users. users.

Nevertheless, this type of attack is rather theoretical, because most smartphones require a series of unsuccessful fingerprint attempts – with Touch ID on iPhone approximately – the entry of a code.

Fingerprints are generated by the IA

However, in DeepMasterPrints, researchers have let artificial intelligence software badyze tens of thousands of fingerprints of real people in a publicly accessible database. In doing so, the software paid attention to structures that recur frequently in actual footprints, and then generated artificial footprints on that basis.

This gave DeepMasterPrints a success rate ten to thirty times higher than randomly generated MasterPrints. When tested with fingerprint sensors with a fault tolerance of 0.1% – according to the researchers, a rate common to the system currently used -, 22.5% of attempts to control the system of fingerprints. security have been achieved. With this success rate, attacks against fingerprint sensors can then be envisaged in practice.

Back to the home page