I'd expect the primary use case to be liveness detection, to validate that the person the facial recognition identifies is not a photo, a sculpture, person wear

I'd expect the primary use case to be liveness detection, to validate that the person the facial recognition identifies is not a photo, a sculpture, person wearing facial prosthetics, or a mask. I've coding such software for that exact purpose.