It is unlikely this will ever work that well in the real world for several very physical reasons.
First, they used an overly ideal four flash lighting setup. This setup was designed not only to properly expose the subject of the photo correctly but also put enough light on the people to be analyzed to be seen in the iris of the subject. For a random photograph the subject would have to be looking at someone that is in a much brighter location than where the subject is located. Granted this does happen but it is only going to be found in a minority of photographs.
Second, sensor resolution is the least important factor in the resolution of an image. The main factors are the lens itself, how far away the subject is from the lens is and what f-stop the aperture (how big the opening) is at. A lens with a smaller opening like those found on a typical cell phone camera will never be able to produce the resolution needed for this technique. This is because of quantum mechanics, the light acts more like a wave than a particle and interferes with itself. This interference always results in blur. The only way around this is to have a larger sensor, that needs a larger lens that needs more physical elements to correct for the spherical aberrations of that lens.
Third is making sure you always have focus on the virtual image reflected off the eyes. A shift in focus of a few millimeter is enough for that image to not be formed at all.