Scientists at the American University of New York have modeled a neural network for counterfeiting fingerprints, putting smartphones with fingerprint sensors in a vulnerable position.
During practical tests, 1/5 of all fingerprints generated by a neural network called DeepMasterPrints could be used on various devices with installed scanners.
Experiment with neural network
In modern smartphones, laptops, tablets, etc. fingerprint sensors for identification of the owner have their own peculiarity - their size is smaller than the fingerprint itself. The device stores multiple files in memory, each of which is part of a full-size print. As a result, a smartphone or other device quickly identifies the owner without the need for repeated scanning from different sides.
The developers of the artificial network DeepMasterPrints took a scanned large database of prints as a basis, subsequently identifying a number of patterns among them. After that, neural networks opened access to the database, which, using machine algorithms, began to fake fingerprint data. The final result of the experiment showed that 23% of fingerprints artificially created by the network can bypass the scanners of mobile devices and sensors at the entrances to various rooms. The researchers intend to improve this figure in the following experiments.
The developers of DeepMasterPrints themselves argue that neural network fingerprinting is a useful practice that helps to identify security vulnerabilities. In the future, the results of the study will become the basis for creating more advanced technologies for protecting personal data. At the same time, scientists do not exclude that a neural network for faking fingerprints can attract attackers interested in obtaining personal information. Thus, the new technology can be not only beneficial but also harmful. For this, hackers do not need a whole fingerprint, they only need to get a copy of a small fragment of it.
The first devices with a fingerprint sensor
For the first time in the world, a fingerprint sensor was introduced in the Atrix smartphone of the famous Motorola brand. At the same time, the technology itself was previously patented by Apple, which presented a method for unlocking a phone using a fingerprint scanner. The first iPhone with such a system was the 5S model, and the technology was named Touch ID.
Despite Apple's claims of virtually zero chances of duplicate areas of different prints, in its 2017 iPhone X and 2018 iPad Pro, the corporation has abandoned this technology, replacing it with a facial identification called Face ID. At the same time, the fingerprint sensor is still present in modern devices of the Apple company, for example, in the 2018 MacBook Air.
Smartphone models with a fingerprint sensor often contain personal information, including financial information linked to bank cards and accounts. For this reason, a neural network for hacking a smartphone casts doubt on the reliability of protecting a phone from someone else's interference. Scientists-developers have not yet shared the technical subtleties of the algorithm for the selection of prints, fearing their use for criminal purposes. Instead, the researchers recommended improving biometric security technologies, making them more resistant to possible hacking.
The Topic of Article: The neural network has learned to fake fingerprints.