The researchers found that to deceive most recognition algorithms, a normal printed mask with the face of another person is enough.
Face recognition systems currentlyused by law enforcement and business as a security tool to keep track of who visits airports, shops, and uses smartphones. However, as it turned out, the program can easily be mistaken.
Kneron, an artificial intelligence company, conducted an experiment on three continents, during which using printed masks were successfully fooledvarious recognition systems. In particular, programs at airports, at border points, passport control, as well as the identifiers of Alipay and WeChat payment systems were mistaken.
However, researchers say that in this wayYou can not get around any system. For example, Apple and Huawei use the more sophisticated technology of structured light imaging, so they successfully passed the reliability test.
Having published the results of their work, the teamKneron encouraged companies and organizations to use higher standards, and developers to improve the quality of their products. According to them, using such manipulations, you can bypass the security systems of payment applications, which potentially allows fraudsters to make purchases on behalf of others.
All tests at the checkpoint, at airports and other public places were carried out with the permission of the security services.
Earlier, we also reported on other methods of cheating recognition systems, which cause algorithms to incorrectly determine objects, the distance to them, or even make them “invisible”.