How to fool facial recognition

[Randy W]

from Abacus

 

True intelligence, no. Intelligently pre-programmed logic, which is definitely fallible.

 

 

Facebook and Huawei are the latest companies trying to fool facial recognition

 

Researchers from Facebook, Huawei and several universities show how software, stickers and clothing can confuse artificial intelligence

 

 

After placing a sticker on the hat, the facial recognition system did not recognize the person as a person. (Picture: Petr Ivanov/YouTube)

 

This video shows "de-id'ed" faces, which are altered enough to break the facial ID logic, but not enough to appear any different to a human eye. You can see the changes (altered eyes, nose, and mouth) using a layered image in Photoshop to switch between the real and the altered faces.

 

 

 

Similar adversarial attacks have been used to trick object detectors. In April, researchers from the University of KU Leuven in Belgium showed how holding a printed pattern in front of them could successfully trick an AI system into thinking that they aren’t human.

 

Some researchers have also printed similar adversarial patterns on T-shirts.

But all these successful experiments don’t mean a single piece of clothing will shield you from all surveillance systems in the world. Many adversarial attacks are designed specifically for one AI system.

So the most effective technique for protecting your privacy against AI surveillance right now is still probably just hiding your face -- like using these “anti-surveillance” coats or applying anti-surveillance makeup. Although these solutions might also make you more conspicuous to your fellow humans.

 

[Randy W]

A Professor, a Zoo, and the Future of Facial Recognition in China
In China, businesses are forcing customers to use facial recognition scanners. But Guo Bing is fighting to outlaw the practice — by taking his local safari park to court.

from the Sixth Tone

Quote

 

In early 2019, Guo had bought an annual pass for a safari park in the eastern city of Hangzhou, so he could take his kid to see the animals on weekends. Six months later, the venue notified him about a new security policy.

In the future, all passholders would have to enter the park using a newly installed face-scanning system, the message stated. Anyone refusing to comply would be denied entry.

Guo, a legal professor at Hangzhou’s Zhejiang Sci-Tech University, was outraged. He couldn’t understand how a zoo could justify requiring customers to hand over their biometric information. When he visited the park, he also found it was doing little to protect the data.

“It was completely messed up,” Guo tells Sixth Tone. “The staff was just using their phones to scan people’s faces. Isn’t it crazy?”

For the academic, the incident was the final straw. He had been growing increasingly concerned about the rapid rollout of facial recognition-enabled cameras in venues across China over recent years, which had proceeded with little accompanying effort to regulate the industry.

He decided it was time to take a stand. In October 2019, he sued the zoo for breach of contract and violation of consumer rights.

The case — widely acknowledged to be the first lawsuit over facial recognition filed in a Chinese court — has generated huge media attention in China, sparking heated discussion over the privacy risks associated with China’s ever-growing use of the technology.

China has been a leading champion of facial recognition in recent years, and face scanners have rapidly become an unavoidable part of everyday life. Chinese citizens are now often required to have their faces scanned when entering train stations, residential compounds, and many other venues. In some cities, even public toilets use the technology.

Though Chinese law prohibits the compulsory collection of facial data, these systems are normally installed regardless of consent and with little prior consultation.

 . . .

Last year, a real estate agent was arrested in southwest China’s Guangxi province on suspicion of stealing over 10 million yuan ($1.54 million) from homeowners by unlocking their accounts on a government-run property app after scanning their faces.

Online, a black market for facial information has also emerged, with packages including an individual’s personal details and a video of their face selling for around 100 yuan, The Beijing News reported.

Guo has become a leading voice calling for greater controls on the use of facial recognition, with his ongoing dispute with Hangzhou Safari Park chiming with growing public skepticism toward face detection systems.

The professor went to court with a modest goal: to force the zoo to end its “overlord clause” requiring visitors to have their faces scanned. By doing so, he hoped to set a legal precedent that would deter other venues from using similar policies.

But the fight has proved tougher than he had imagined. Judges in Hangzhou have been reluctant to rule on the wider issue of whether the zoo’s use of facial recognition is justified, instead focusing narrowly on the individual contract dispute between Guo and the zoo.

After the first trial, the court ruled in Guo’s favor, but simply ordered the safari park to delete the professor’s facial data and pay just over 1,000 yuan in compensation. Guo appealed, and a second trial was held in Hangzhou Intermediate People’s Court this past December.