What is Face ID and how does it work on iPhone X?

Face security built into Apple's new phone comes with a claim that the confusion rate is only one part per million.

What is Face ID?

Face ID is a biometric security feature based on face recognition, which Apple introduced into iPhone X to replace Touch ID fingerprints. The technology is introduced with high reliability, convenient usage, and comes with a range of new equipment that is partly making the iPhone X expensive. So how can the machine determine the identity of the owner?

"The device will compare two faces and identify similarities," said Professor Anil Jain of Michigan State University (USA). "The easiest way is to record the points on your face. You will have to 'mark' your face, like fingerprinting with Touch ID."

Take photos of faces

TrueDepth camera components help Face ID on iPhone X work.(Photo: Apple).

Face ID uses a cluster of lights, combined with sensors to record the owner's face. This system is called by the name of the TrueDepth camera by Apple and the company says it will "create a detailed face depth map to be able to recognize you immediately".

In order to set up Face ID, users follow Apple's instructions, including an antique rotation so that the TrueDepth camera can identify faces at different angles to build a 3D map. The system uses infrared (IR) to illuminate the face, helping to identify even in the night, regardless of indoor or outdoor. IR is electromagnetic radiation with a longer wavelength than visible light.

"Infrared lights emit light that the human eye cannot see, it will support in low light conditions, or compensate when the environment light is not favorable , " Jain explained. "So the TrueDepth system allows facial recognition even when it's too dark or too bright."

Infrared light system on iPhone X when operation is recorded by infrared camera.(Photo: Verge).

For face recognition, first the proximity sensor and light sensor help the TrueDepth camera determine the brightness of the environment needed for face verification. Next, the flood illuminator system will emit infrared light to illuminate the user's face.

A dot projector emits more than 30,000 invisible lights to build a three-dimensional map, including the depth of the entire face area. An infrared camera will capture pixels and infrared light reflected from your face.

Face recognition

The face is one of biometric elements - a measurable biological property. This is considered an effective personal identification tool, in addition to the face can include fingerprints, iris or even voice .

Try Face ID feature on iPhone X. (Video: Dinh Nam).

Basic biometric authentication systems all work by calculating and comparing the two models to see how they are identical. These patterns can be waveforms in the voice, lines on the fingertips (hand flowers), iris structure or facial features.

When setting up a biometric system, a computer - such as a phone processor - will capture and store a reference, called a form or "registration form". Then, when the user wants to access the device, such as unlocking the phone, you will have to give "verification image" to the computer.

"The computer system will calculate with two states, 0 and 1 , " Jain said. "If the image matches close to 1, this means the fingerprint or face is the same. If it is near zero, synonym verification is not the same as the original, not the same person."

Since registration images and verification images will not be the same due to different shooting conditions, the phone will use a threshold to determine if they are significantly similar. For example, the comparison score reaches 0.7, it can be considered that the verification is identical to the original, but this number is not fixed.

Face ID activity on iPhone X. (Video: BI).

"If you only unlock the phone, the threshold set by the manufacturer may be relatively low, such as 0.5-0.6," Jain added. In some contexts, this number may be higher, such as when you pay for your mobile wallet. "If this is a transaction worth $ 10,000, this threshold may rise to 0.9".

All calculations happen very quickly thanks to the processor inside the device. For example, iPhone X with "neural engine" (Neural Engine) can perform 600 billion calculations per second and it is used to process Face ID face recognition in real time.

Thus, to confirm the face on iPhone X, the infrared image at the shooting step will be sent to the iPhone X processor to build 3D "map" of the user's face, then present it under format of computer algorithms. The device will match the stored pattern (registration form) and calculate how many points it will earn. The machine will authenticate the identity and unlock it when the comparison point is higher than the threshold value.

Accuracy when authenticating

There are three factors that affect the accuracy of authentication: pose (pose), illumination and expression.

"Under ideal conditions, if users cooperate, face-to-face, uniform lighting and normal expression levels, the accuracy of authentication can reach 99.99% , " Jain said. Obviously the authentication will be more accurate when your face is not covered by anything, such as sunglasses or scars.

"With personal papers such as passports, visas, driving licenses . you are required to provide standard images , " Jain added. "You must not laugh or smile at passport photos, the background must be white and of course remove any accessories on the face".

Not limited to ideal conditions, Apple claims Face ID can work regardless of user appearance changes, such as when you wear glasses, cut hair or shave. This capability is achieved by Neural Engine technology, which iPhone X has a "specialized hardware" to "learn" situations and Apple also develops many "neural networks" for processing.

Touch ID introduced by Apple with 50,000 part accuracy, the advertised Face ID is only one part million. The company is also confident the new technology can identify faces in most cases, even if users wear glasses, make-up or change hairstyles . and are not overcome by masks.

Some experiments show that Face ID may be "fooled" by twins (but not every pair is successful). WSJ and Wired have invited some experts to create masks costing thousands of dollars, with a variety of materials such as silicon, gelatin or plastic . to make up the eye area, eyelashes . but still not "penetrated" this security barrier of Apple.

Bkav uses a mask to fool Face ID.(Video: Bkav).

Meanwhile, Bkav launched a video claiming they could make a Face ID trick on the iPhone X. Notably, the Vietnamese company said the mask only cost about 3.4 million to complete."It's even simpler than we thought," Wired leads Bkav's information.

Bkav's results were noticeable in the technology village when it was the first company to declare success in deceiving Face ID. However, independent security experts have yet to speak out about the authenticity of Bkav's video. At the same time, the tech community raises questions about the process that this security company has done by video has not yet made it clear.