It uses a depth sensor. And they have trained the neural networks to prevent using 3d printed masks instead of the real face. I guess they are using depth-ir-po

It uses a depth sensor. And they have trained the neural networks to prevent using 3d printed masks instead of the real face. I guess they are using depth-ir-powered-video instead of still photos.