Gadgets

Facebook taught AI to cheat face recognition systems (video)


Artificial System Developersintelligence is currently working on a methodology to counter the spread of fakes and increasingly high-tech tools used to create them. Facebook, more than once involved in scandals with violation of privacy, offers a methodology to combat fakes, in particular with the formation of fake video.

Facebook Specializing ina study of high technology Facebook AI Research (FAIR), has developed an algorithm that allows you to deceive the face recognition system. The available technologies, so far, have offered only a detour of tracking and personality recognition systems using photographs of a static object. FAIR experts offer a technique for misleading video systems. The algorithm allows you to cheat even in real time during live broadcasts.


The FAIR algorithm is based on a techniquealmost invisible to the human eye changes in facial features with the help of machine learning. However, such changes will not allow the face recognition system to uniquely identify a person.

Face Recognition May Loseprivacy, and face-swapping technology can be used to create misleading videos. Algorithms developed by FAIR are designed to deal with such cases of technology application.

It should be noted that in Russia, in Moscowlawsuit on the suit of a resident of the capital against the authorities on the legality of the use of identity recognition systems in urban video surveillance systems. According to the claimant, this technology violates the right of Russians to privacy guaranteed by the Constitution. All biometric data of citizens can be used only with their written consent.

Source: venturebeat