Let’s say you live in a country where there’s been reports of massive outbreaks of a highly contagious viral infection. Hypothetically, let’s call this infection “coronavirus.” To avoid spreading this “coronavirus,” many people have taken to wearing medical face masks when going about their everyday life. These masks don’t do much, medically speaking, but it’s considered the polite thing to do.
Now, let’s also say that the government there is notorious for widespread and unregulated use of facial-recognition technology as a way to both fight crime and to identify and silence political dissidents.
This government has taken to banning face masks at massive anti-regime protests that have been shaking one of its more economically viable, semi-independent territories for the better part of a year, in an effort to better identify and target the protesters.
So, what would this government do, if suddenly most of its population begins wearing face masks on the regular?
Well, if you’re China, you might be seriously considering the government grant application of one Megvii, a Chinese unicorn startup known for developing the Face++ facial-recognition system. According to Reuters, Megvii has applied for a 100 million yuan ($14.32 million U.S.) grant to improve technology that can identify people who are wearing masks in crowds.
“Regardless of what you think the intentions are, it’s obvious what this kind of tech is for,” said Chris Kennedy, chief information security officer at AttackIQ, a cybersecurity for business firm. “If a masked robber commits a crime, you can track them down quickly. But when does a person become criminal because they’re dissatisfied with their government? This can be used against you.”
Kennedy said it was “telling” that Megvii was asking for the grant. “The pollution problem in China is real. The biological and virus problems are real. People will start taking precautions,” he told Digital Trends. “If there is a broad government plot to lay out surveillance infrastructure, all of that has just been upended by the fact that it’s unhealthy to live there. This exposes that they’re already thinking forward.”
In an email statement to Digital Trends, a spokesperson for Megvii denied that the new solution had anything to do with identifying people with masks. “It only screens people to measure their body temperature even if they are over three meters away and wearing masks, and then locate those among the crowd that have elevated body temperature, so that ground staff can double check their temperature with traditional thermometers,” they wrote. “The solution cuts close physical contact between ground staff and passengers passing through the subway station, speed up passenger flow and cut or avoid long queues, so as to reduce the risks of transmission of the coronavirus.”
Experts say any Hong Kong protester, many of whom have been in the streets calling for a more democratic and independent Hong Kong for almost a year, should be worried.
The U.S. blacklisted the company in October 2019, during the initial public offering (IPO) in Hong Kong, for alleged participation in human rights abuses of Uighur Muslims in Xinjiang; Megvii ultimately cleared $500 million in IPO funding in January 2020.
Attila Tomaschek, digital privacy expert at the U.K.-based ProPrivacy, told Digital Trends in a statement that this is “a downright terrifying glimpse at a dystopian future devoid of any and all dissidence.”
“Assuming the project is supported, it would show that Beijing will stop at nothing to exert its authoritarian control over its citizens,” Tomaschek wrote. “The prospect of such advancements to facial-recognition technology is certainly alarming and something those on the ground in Hong Kong – and around the globe – will be keeping a close eye on.”
This kind of algorithm — being able to identify people based on just part of their face— is extremely feasible, if not quite as accurate as seeing someone’s full face, said Labhesh Patel, chief technology officer and chief scientist at Jumio, a online ID-verification company.
“A.I. can still do quite a bit of matching with just eyes and a forehead,” he told Digital Trends. “What will come up in a database is maybe five or ten pictures, instead of one or two if you had the whole face. So it is feasible to narrow down a search space, and then use other modalities [cell phone geolocating, for example] to locate a person.”
Is the facial-recognition genie out of the technological bottle? Or can it be reined in? Maybe, if regulators do something, but probably not in China.
“The reality is, we all know that China has not been one to follow any internal regulation,” said Kennedy. “Most of that is a facade to pacify their more liberal constituents.”
“I think regulatory bodies need to get their act together,” Patel said. “Right now, this is the Wild West of facial recognition. It’s a completely unregulated environment. There are no repercussions for companies like Face++.”
It’s a tricky balancing act between using facial recognition for personal security, for good causes, or for nefarious surveillance purposes, Patel said. He pointed to the story in India where the government set up a database of missing children and, through facial recognition, was able to locate several hundred of them.
“It’s a compelling use case,” he said. “It’s also in the same realm of use cases as trying to find a protester. Some of these things, I believe, can only be solved through better regulation.”
While new regulations like the General Data Protection Regulation in Europe and the California Consumer Privacy Act were steps in the right direction, Patel said there is frustration in U.S. academic and business circles that China seems to be nothing but “a black hole that sucks research from everywhere else,” but doesn’t provide any output. “I don’t think there are any steps being taken to address these issues,” he said.
Correction 2/13: An earlier version of this story referred to a Human Rights Watch report that alleged Megvii’s involvement in a mass surveillance app monitoring Uighurs in Xinjiang, China. That report has been corrected, and the organization “has since confirmed that the Face++ code in the IJOP app, which was in the log-in function, was inoperable.”
- Federal bill would ban corporate facial recognition without consent
- Ring may want to add facial recognition and more to cameras
- UCLA won’t use facial recognition on its campus after it receives backlash
- U.S. military facial recognition could identify people from 1 km away
- This invisible photo filter protects you from facial recognition software