It is regularly contested and criticized as much as, on the contrary, considered as the ultimate identification solution. What is certain is that facial recognition uses extremely sophisticated algorithms. Decryption.
An alternative to fingerprint sensors, facial recognition unlocking technology is present on more and more smartphones. The iPhone X of course, but also the latest Galaxy S from Samsung, the OnePlus 5T and the mobiles of the Nexus and Pixel family from Google. With more or less advanced technical solutions and varying success.
Mapping the face
These devices are based on a simple principle: the comparison between a reference image, the imprint recorded by the user of the terminal, and the face filmed by the camera lens. Algorithms take measurements on the key characteristics of the human face: the distance between the eyes, the edges of the nose, the corners of the lips, the ears, the distance between the chin and the lower lip, etc. Taken end to end, these data make it possible to define a profile which is intended to be specific to each individual.
Analyze the relief to improve identification
Face recognition devices implemented on smartphones rely on two different techniques. The first concerns 2D recognition and is quite easy to deceive with a simple paper photo. The use of the motion sensor sometimes completes the device and prevents it from being deceived in this way. The second technique, called 3D recognition, consists of projecting light marks (up to 30,000 beams on Apple’s TrueDepth) on the user’s face. Then to analyze their deformation to draw up a three-dimensional map, much more reliable than a 2D footprint. This information is encrypted and stored on the phone.
In search of reliability
The potential of facial recognition is undeniable. It is much more natural to look at the screen to unlock your phone than to put your finger on a fingerprint sensor. However, this process comes up against the technical limits of current devices. Smartphones that have it are still not able to process enough information to form a truly reliable faceprint. Their chips also lack the power to analyze in a few hundredths of a second the face that appears in front of the lens to unlock the mobile. To improve the efficiency of facial recognition and bring reliability to the level required for online payment validation operations, manufacturers are working on complex sensors (measuring distance, movement, etc.) and chips with motors. neurons capable of carrying out hundreds of billions of operations per second and thwarting traps: distinguishing between parents and twins, managing changes in appearance – beard, glasses, haircut, cut … ‘a matter of time.