iPhone X: Apple explains how new phone's facial recognition system keeps users' faces safe

Apple head of marketing Phil Schiller introduces the new iPhone X during an Apple special event at the Steve Jobs Theatre on the Apple Park campus on September 12, 2017 in Cupertino, California (Justin Sullivan/Getty Images)

Apple has revealed detail information on how the controversial facial recognition tool on the iPhone X works.

The feature is intended as a safe, secure way of quickly unlocking a phone without even touching it. But the way it works – creating an incredibly detailed map of your face, then storing that away – has the potential not only to be strange but outright dangerous.

Many of those objections were raised earlier this month, when Apple showed off the product. Though it took a range of steps to make clear that the phone would be safe from people using tricks like photos or masks, and the data it stores would be kept entirely private, questions have remained about whether the biometric scanning system can be trusted.

The company has published a detailed white paper that sets out precisely how Face ID is built, to show how it is safe. It doesn't explain all of the technology behind it – much of which is secret and belongs to Apple – but makes clear how the phone ensures that the detailed information it collects is kept secure, as well as how the phone is kept safe.

The facial recognition data never actually leaves the phone, for instance – and even inside of the phone, is kept in a specifically secure part, meaning that even Apple can't see it.

Other apps can use the facial recognition technology. But they never actually see any of that information in that case, either – they simply ask the phone whether the person using it is the owner, and receive a yes or no back.

That said, some apps can use the information generated by some of the same cameras, though. On stage, Apple showed off how Snapchat can use the facial mapping to generate new kinds of filters on your face – but in that case it only receives a "mesh" picture of your face, in nothing like the right amount of detail to be able to use it to trick its way into your phone.

Apple also makes clear that the phone adds to its mathematical representation of your face over time. So if, for instance, you wear glasses and then look at the phone, it might not initially recognise you – but once you get your way in with a passcode, it will add the image of you with glasses on and be able to look out for that in the future.

And an entirely extra neural network is watching all the time to see whether something strange is going on, and if someone appears to be trick the phone – using things like a photograph, which couldn't work because the phone needs to map the face in 3D, or a more sophisticated map. If it sees that and thinks something is wrong, then it will lock down the phone and allow it only to be opened with a passcode.

That is the phone's ultimate fallback in all cases, just like it is with TouchID, so that people can always switch to only using a passcode that's in their head. And Apple offers a new way to trigger that option, too – a new, so-called "cop mode", which allows people to press a combination of buttons and turn off TouchID or FaceID and require a passcode.