Windows 10's Hello feature is facial recognition technology that replaces passwords and is available on some newer compatible devices. An experiment involving identical twins indicates that the system is accurate, as it could distinguish between the very similar-looking siblings.

An Australian journalist put the new technology to the ultimate test by recruiting six sets of identical twins to determine whether Windows 10 and a compatible device could distinguish between the siblings. The device used was a Lenovo ThinkPad Yoga 14, equipped with the necessary Intel RealSense camera. The camera uses an infrared lens, a regular lens and a 3D lens to utilize photographic recognition, depth and heat detection in creating its facial analysis.

The heat sensor prevents hackers from using photographs of the real deal, while the IR lens helps adjust for variations that can result from facial hair, cosmetics use and different lighting conditions. The experiment began with one person from each set of twins registering for a Windows account while utilizing the face recognition process. The process allows for greater accuracy if the user registers more than one image to account for different looks they might sport — for example, with and without eyeglasses.

The technology was very successful at differentiating between the twins. Out of six sets of twins, the system managed to recognize the correct person on all but one occasion, in which the computer recognized neither of the twins in a set. In no instance did the technology allow an unregistered twin to access the computer on which his or her sibling had registered.

Annabelle and Miriam Jeffrey were two of the twins whom the computer successfully differentiated.

"It could distinguish between us two quite easily," said Miriam. "It's a little surprising; I thought it would have failed, but no, it was really good, it was really quick."

Microsoft claims the system has a false positive rate of less than 1 in 100,000.

ⓒ 2021 All rights reserved. Do not reproduce without permission.