Cheating security cameras with a painted face

  • A team from campaign group Camoflags wanted to know if special face paint designs could fool facial recognition cameras.
  • A software program that works on the basis of artificial intelligence has created a series of designs that are particularly suitable for football fans.
  • The team found that facial recognition software can be cheated in a number of cases.
  • Read also: How to ensure as an organization that you are responsible for artificial intelligence and not the other way around

What if you could paint your face in such a way that it fooled facial recognition cameras at the World Cup in Qatar?

Action group Camoflags sought the answer to that question. It used designs to paint your face generated with software based on artificial intelligence. The designs are based on the make-up of the famous hip-hop group Insane Clown Posse: Juggalo.

The experiment is ideally suited for the World Cup in Qatar, as many football fans paint their faces to support their country.

Face painted with the American flag. Photo: Camoflags

The World Cup in Qatar is criticized for many things, but also for its security policy. There are serious concerns that the tournament is a breeding ground for espionage and that participants are being tracked through the mandatory apps on their phones.

The tournament has already received some criticism leading up to it for human rights violations. Around 15,000 facial recognition cameras are being deployed in and around World Cup stadiums to monitor terrorist threats and hooligans, the World Cup chief of technology told AFP in August.

“What you see here is a new standard, a new trend in event security. This is our contribution from Qatar to the world of sports. This is the future of stadium security,” Niyas Abdulrahiman told AFP.

Tao Thomsen of Camoflags, a project of the creative agency Virtue, tells Insider that his team came up with the idea years ago to test facial recognition with strange makeup. The project initially focused on tricking deepfake AI technology into preventing women’s faces from being used in sex or other videos without their consent.

Painting with the Tunisian flag. Photo: Camoflags

The idea of ​​using makeup to trick facial recognition algorithms is not new. In 2010, artist Adam Harvey’s CV Dazzle project was one of the first attempts to trick artificial intelligence with geometric shapes.

There is also evidence that hip-hop group Insane Clown Posse’s Juggalo makeup is ideal for fooling facial recognition systems.

The problem with such projects was the practical application, says Thomsen. His team couldn’t ask women to wear weird makeup all the time. But in football it is different.

“Football games are the perfect occasion to paint strange patterns on your face,” says Thomsen. “People expect you to paint the colors of your country and your flag on your face.”

Painted with the South Korean flag. Photo: Camoflags

To arrive at the ideal design, Camoflags fed the artificial intelligence software Midjourney AI with descriptions so that designs with a perfect combination of Juggalo makeup and various flags would roll out.

“We went through the whole design process many times until we got a result that the facial recognition software didn’t recognize as a face,” says Thomsen. “And afterwards, some human love had to be involved to make the patterns somewhat aesthetic, because generative AI starts hallucinating at some point and then starts doing really crazy things.”

The team tested the design on different types of facial recognition technology and found that many of the designs could fool cameras. It is not known what camera system is currently used in Qatar.

Camoflags estimates that 80 percent of tests were successful with ten systems tested.

The Brazilian flag. Photo: Camoflags

Camoflags has yet to see soccer fans in Qatar using their designs, but believes the project will help spark a discussion about surveillance at sporting events.

According to Thomsen, British football fans have already experienced facial recognition errors during the 2017 Champions League final in Cardiff.

Stadiums around the world are now using facial recognition technology to identify people involved in criminal activity, collect data for advertisers and recognize ticket holders at events.

There is evidence that facial recognition technology works less well with women and people of color, which can falsely link people to a crime.

READ ALSO: World Cup in football in Qatar: these are the 9 most expensive teams – Holland is included

Leave a Comment