This 'tech' is just a theoretical lab exercise with zero real world applications.
Basically they had a router in one corner of a room and receivers at various points on the other sides of the room. And cameras. When people walk around in the room they block some of the signal and with machine learning they linked the patterns in interference to the images of people standing in certain positions. Once you have enough of this data (possibly months) you can turn off the camera and the software will recall interference pattern 7547785541 to match frame 10093439458 of the stored imagery,
Now if some hacker installed this software remotely on your router it does absolutely nothing as there are no receivers to measure. And if they snuck in and hid receivers they still have no reference data to work from. Their test room is unlikely to have the same router position, furniture and dimensions. So they need to sneak in a bunch of cameras as well. And if they did that why do they need this router radar.