AI-designed makeup could fool facial recognition cameras at World Cup

  • A team from Camoflags wanted to see if face paint designs could outperform facial recognition cameras.
  • The design, which uses Juggalo makeup concepts, was created by an AI program.
  • The team found that many designs could circumvent various facial recognition software.

What if a simple face paint could fool some of the best facial recognition tools available at the Qatar World Cup?

A team known as Camoflags sought to answer that question with AI-generated, Juggalo-inspired facepaint designs that could be used to evade facial recognition cameras.

The experience is particularly suited to the World Cup, as face painting is a common feature at football matches for fans who support their teams.

A person with red, white and blue face paint

American flag face paint stylized as Juggalo make up

Courtesy of Camoflags

Qatar’s World Cup, which ends on December 18, has been criticized for its approach to security, with fears the event could become a hotbed of espionage and visitors could be monitored on their phones through surveillance apps.

The event has also been criticized for its human rights violations: the death of American sports journalist Grant Wahl, who died on the way to hospital after collapsing at the World Cup stadium, sparked a renewed focus on LGBTQ rights in Qatar. Wahl had previously been arrested by authorities outside the World Cup stadium for wearing a rainbow shirt.

Around 15,000 facial recognition cameras have been installed in and around the stadium to identify terrorist threats and “hooliganism”, the World Cup’s chief technology officer told AFP in August.

“What you see here is a new normal, a new trend in venue operation, it is our contribution of Qatar to the world of sport. What you see here is the future of venue operation. stadiums,” Niyas Abdulrahiman told AFP.

Tao Thomsen, one of the team members of Camoflags, a project by creative agency Virtue, told Insider that the team came up with the idea of ​​testing facial recognition with strange makeup combinations that distorted facial features years ago in response to growing popularity. deepfakes of women created without their consent. The original intention of the project was to create makeup looks for women that deep AI technology could not recognize, thereby protecting their faces from being used in sexually suggestive or non-consensual videos.

A person with red and black face paint

Tunisian flag face paint stylized in Juggalo makeup

Courtesy of Camoflags

The idea of ​​using facial makeup to try to fool facial recognition algorithms is not new. CV Dazzle, a 2010 project by artist Adam Harvey, was one of the first attempts to use geometric shapes and other gimmicks to outsmart AI. There’s also evidence that Juggalo makeup — which is clown-like makeup inspired by the hip-hop group Insane Clown Posse — in particular, is effective at tricking facial recognition systems.

The challenge with projects like these was their practical application, Thomsen said. The team realized that it would be difficult to ask women to wear “crazy” makeup combinations all the time. But with football it was different.

“Because if there’s one place where it’s okay to have weird patterns on your face, it’s a [soccer] game,” Thomsen said. “People expect you to paint your team colors and flags and all that on your face.”

A person with white, red, blue and black face paint

Face painting of the South Korean flag stylized as Juggalo make up

Courtesy of Camoflags

In order to create the ideal shape, Camoflags fed an AI art software tool, Midjourney AI, a series of descriptions that produced designs with the perfect mix of Juggalo makeup and various flags.

“We gave it these kinds of descriptions and then ran through this process, many times until we got a result that couldn’t be recognized as a face when we passed it through facial recognition software,” he said. Thomsen said. “And then we knew we had some human love afterwards to tidy things up and make sure that [the patterns] are also aesthetically functional because the generative AI starts hallucinating at some point and doing really weird stuff.”

The team tested the different Juggalo and flag-patterned make-ups on different consumer-grade facial recognition technologies – the face camera system used in Qatar is unknown to the public – and found that many models could evade cameras.

Of the 10 systems tested, the team estimates that around 80% of the tests passed.

A person with green and yellow paint

Face painting of the Brazilian flag stylized as Juggalo make up

Courtesy of Camoflags

Although Camoflags is not yet aware of their face paint designs being adopted at the World Cup – and recommends people not risk getting in trouble in Qatar by testing its limits – they think it will help. start a conversation about surveillance at sporting events.

Another member of the Camoflags team, Morten Grubak, told Insider that the team had created face filters that would allow users to place virtual Juggalo designs and “color them in a bit like a drawing book.” using face paint.

Thomsen specifically pointed to the UK, where football fans have already had to deal with faulty facial recognition. Sports venues around the world have also started using facial recognition technology to identify people engaging in illegal activities, collect data for advertisements and verify ticket holders for entry at events.

There is also evidence that facial recognition technology is less likely to correctly identify women and people of color, which could lead to misidentification of people for crimes by law enforcement.

“We know that there is a large part of [soccer] fans, especially in the UK, who are very clear about their mistrust and aversion to facial recognition, and we hope this will become a tool for [soccer] fans there and around the world, to take a stand against facial recognition,” Thomsen said.


Not all news on the site expresses the point of view of the site, but we transmit this news automatically and translate it through programmatic technology on the site and not from a human editor.
Back to top button