The New iPhone Isn't on Sale Yet, but the Privacy Anxiety Over Facial Recognition Has Already Begun

罗骢 · · English translation

By now, you can't do without the technologies that make your life easier and may also give your privacy away.

By now, you can't do without the technologies that make your life easier and may also give your privacy away.

In 15 days, the iPhone X goes on sale.

On buildings and in train stations, iPhone 8 billboards that went up less than a month ago have already been replaced with photos of the new phone. Apple CEO Tim Cook says it will "define the next decade of smartphones."

The biggest change is the facial recognition feature, Face ID.

More than anything else, the face is how humans tell one another apart. Babies four months old can already read facial expressions, telling a smile from anger.

We use our faces to show joy, anger, sorrow and delight, and to leave a first impression others can't shake.

As the American philosopher Ayn Rand put it, with some exaggeration: "Nothing is as significant as a human face. Nor as eloquent. We can never really know another person, except by our first glance at him. Because, in that glance, we know everything. Even though we're not always wise enough to unravel the knowledge."

Machines still can't truly read a person, but the technology is now good enough for them to recognize one.

"A computer could never do that the way a person can…" — one more human privilege, gone.

That is probably why, even though your privacy has long since been scattered among merchants, institutions, hardware and software companies and internet services, an announcement from Apple — which makes 200 million iPhones a year — still set off anxiety and panic.

Facial recognition is not a new technology, but Apple's version makes mass surveillance possible

Three days after the iPhone X launch, Tim Cook received a letter of inquiry from US Senator Al Franken, chairman of the Judiciary Subcommittee on Privacy, Technology and the Law, asking whether Apple would use facial data as it pleased, and how it would protect that data.

On October 17, Apple replied formally, repeating that facial data would be stored only on the phone itself, explaining that its samples were not skewed toward white faces, and noting that Face ID had been trained with AI so photographs or other models couldn't fool it.

This is not the first time Apple has faced questions about privacy. In 2013, as soon as fingerprint recognition appeared on the iPhone 5s, people asked whether the data might leak. On Twitter, some even tried to mail Apple their own hair along with a 5s, because hair contains DNA.

Facial recognition is not, in fact, a new technology. Anyone who follows tech news is unlikely to have heard of it for the first time at the iPhone X launch.

Internet companies' staff-free convenience stores. Banks' face-scanning ATMs. Even many phone apps that handle money are already recognizing your face.

Governments haven't lagged behind. In Britain, retailers use it to spot customers with shoplifting convictions. In Russia, Moscow has built a surveillance network that tracks faces. In China, a system called Skynet is quietly in development.

Alipay's Smile to Pay

The big problems with recognition right now are efficiency and accuracy.

Facial recognition products from Alipay and other companies can already identify a face reliably, but they all need you to look straight into the camera. Some, such as ATMs, even ask you to move your head as the screen instructs.

A city's surveillance system can't count on always seeing a face head-on.

The most effective tool for that is still the human brain. British police have a special unit known as "super-recognizers" — officers who stare at photographs all day and find suspects far faster than AI does.

These limits come from identifying faces in 2D images.

Face ID takes a different approach: it projects more than 30,000 infrared dots onto your face, photographs them with an infrared camera, builds a detailed model and recognizes the face.

Apple demonstrating the 30,000 points Face ID reads

The technology first appeared in Microsoft's Kinect; the company behind it, PrimeSense, was bought by Apple in 2013. It does not really understand a face, but its processing speed and dedicated hardware let it tell faces apart faster than a person can.

It can even see through sunglasses, because most sunglass coatings don't block infrared light.

Putting this into a phone means it won't be long before it can be installed cheaply almost anywhere — at subway gates, for instance, photographing and identifying people at the ordinary pace of a turnstile.

Where you are is no longer a secret. In the science-fiction classic Minority Report, the cameras covering the city could identify people only by their irises. Now a phone costing less than 10,000 yuan does it faster.

A face is public information, but it can give away far more

Besides privacy in public space, phones themselves are now recording people's facial features.

Ming-Chi Kuo, the KGI Securities analyst who knows the phone supply chain well, said recently that since the iPhone X was unveiled, the number of Android phone makers asking about 3D facial recognition has at least tripled.

Which means unlocking a phone may switch over to the face entirely.

Cameras can only watch and record in public places; phones are far more numerous and far more densely spread across the world. The mobile market research firm Newzoo's annual report says that by 2022 the world will have 3.6 billion smartphone users, half its population.

Our faces are becoming data, like fingerprints — code that can be recognized quickly and stored in bulk.

Every industry has a reason to embrace this. Take advertising. If WeChat, Weibo or Facebook could learn which places you go to in the physical world, they could aim ads at you inside their own services.

At these companies, most engineers' work points, in the end, toward more precise advertising.

Fingerprint collection makes people suspicious. Faces are different: a face belongs to one person alone, and yet it is public.

On social networks, hundreds of millions of photographs with faces in them are made public every day.

Even political acts such as marches will no longer stay private. When someone's photo turned up in a protest march, perhaps only their friends could pick them out. Now that information is collected in bulk and at speed, identifying a political protester takes no effort at all.

The second question in Senator Franken's letter was about how Face ID encrypts and stores the facial data its sensors collect.

On the iPhone X, Apple uses differential privacy to protect facial data, adding a great deal of "noise" so that the data about any one person is blurred — as if hundreds of guitarists were all strumming at random at once, making it hard to pick out the one who is really off key.

But most people are not using Apple's phones. And phones are not the only things that recognize your face.

Once the technology is everywhere, leaks of facial data will be as inevitable as today's leaks of email passwords or hotel credit card records.

Governments as well as tech companies have started collecting information with facial recognition, and because they are held to account less often, they generally keep data less securely than companies do.

Just from recent examples abroad: the Australian military leaked a large amount of defense information through badly set passwords; Sweden's Transport Agency had its entire database stolen after being outsourced through several layers of contractors.

When your information leaked before, you could still switch email addresses or change your password. Changing your face is harder.

How it's collected, how it's used: facial data has no privacy boundary at all right now

Alvaro Bedoya, executive director of the Georgetown Center on Privacy and Technology, once asked: "If someone walks down the street, shouldn't a company with no connection to them have to ask their permission before identifying them?"

It isn't only tech companies. More and more governments are announcing surveillance systems with facial recognition built in, and you could be marked without knowing a thing.

The latest is Australia. On October 5, the federal government said it would require states and territories to hand over photos of driver's license holders to build a national facial recognition database, and would let the government instantly scan the faces in footage or images taken by closed-circuit cameras, to identify people and pick out possible criminal targets.

Australia is not alone. Russia announced that since February this year it has used facial recognition on 160,000 closed-circuit television cameras across the country.

China's government is installing large numbers of cameras too. Public figures from 2015 showed that urban Beijing had reached 100 percent surveillance coverage, with more than 470,000 cameras — more than any other city.

Governments everywhere are also pushing the technology hard. The Economist reported that the services sold by Megvii and other Chinese facial recognition companies depend heavily on huge quantities of good data. They have access to a Chinese government database of images of 700 million citizens, including the ID photograph each person had taken at 16.

Much of the time the technology is used on things that hardly matter. In Beijing, the city authorities have used it to identify people stealing paper from public toilets; the system allows each person 60 centimeters of paper every nine minutes. Facial data for 60 centimeters of toilet paper.

Scanning a face to get toilet paper

A contract published by the British government shows that police in Wales planned to use facial recognition on fans at a football final. The faces of 100,000 fans would be scanned and stored at train stations and at the stadium gates.

Eric Schmidt, Google's former chief executive, once said: "If you have something that you don't want anyone to know, maybe you shouldn't be doing it in the first place."

That sounds like a line from a Dave Eggers character. This way of thinking could one day cost us the last of our privacy.

Should facial data be used without asking the people it belongs to? There is still no clear agreement.

European regulators have written a set of principles into data protection rules due to take effect: biometric information, "faceprints" included, belongs to its owner, and using it requires that person's consent.

Britain, which has more cameras than anywhere else, has this rule: you can ask for a copy of the surveillance footage that shows your face, but you cannot have the original deleted.

Policy still can't keep up with the technology, and facial recognition keeps getting better.

The irony is that we hand over more private information in order to be safer

Fingerprints or faces, the point of these privacy technologies — and the reason they spread — is safety and convenience.

On the first iPhone, the default setup didn't ask you to set a passcode at all. As late as 2014 the passcode was still off by default, and setting a four-digit code took a tedious series of steps.

But as the mobile internet grew fast over the past few years, phones took on far more than the texting and music playback they used to handle.

You keep photos and documents on your phone, you pay and shop with it, and your social networks hold your contacts and your relationships.

Those functions need encryption and unlocking that are easier to use. It's hard now to picture 2007, when the first iPhone opened with a swipe and needed no encryption at all.

For most people a fingerprint is far safer than a passcode. By one count, the most common four-digit unlock codes are "1234," "0000," "2580" and "0852."

Because phones matter so much, society has come to accept trading a fingerprint for security — something once unimaginable. After 9/11, when the United States required travelers entering the country to leave fingerprints, protests broke out around the world.

For most ordinary consumers, turning a fingerprint into data adds only a privacy risk that looks remote, and buys the convenience of unlocking a phone countless times a day.

Facial recognition is the same: most of the time, all you will notice is the convenience.

Craig Federighi, Apple's head of software engineering, said in an interview: "Just like when Apple first introduced Touch ID, a lot of people thought back then that they would never use the feature. But now everyone's life relies on Touch ID — and it will be the same with Face ID."

Cover image: The Economist

Translated from the original QDaily article. Original publication date retained.