3D masks manage to trap facial recognition cameras

3D masks manage to trap facial recognition cameras

Technology

An artificial intelligence company recently conducted an interesting experiment. The objective? Trap facial recognition technologies in various locations using high-quality 3D masks.

Fool this technology
As everyone knows, China is renowned for its massive use of facial recognition. the latest application has been unveiled in Shenzhen, a large city close to Hong Kong. According to a South China Morning Post article published in September 2019, the Shenzhen city subway has adopted facial recognition. The objective? Allow users to use this bias to pay for their tickets.

According to an article in Fortune on December 12, 2019, a company named Kneron based in the United States and Taiwan wanted to conduct an experiment. This company specialized in artificial intelligence (AI) simply wanted to trap facial recognition cameras, a business having been crowned with success! Using high-quality 3D masks and simple photographs, Kneron members have trapped such systems managed by the Chinese giants AliPay and WeChat. Another step was to trap a camera located in one of the terminals at the main airport in the Netherlands (Schiphol).

Real risks
The shortcomings of this technology have been proven once again. The most alarming seems to be the fact that this is a means presented as being very sure to prove and validate the identity of a person! Evil people could obviously use this kind of trick to trick the system and pretend to be someone else. And what about the risks possibly linked to terrorism?

Kneron CEO Albert Liu spoke about the experiment. According to him, facial recognition technology could obviously be very useful. However, the companies that currently operate it are not sufficiently secure.

In April 2019, Belgian engineers had already blamed an image recognition system using simple signs. The objective? Show that even fairly simple means can be enough to trap algorithms considered to be very powerful.