I think he forgot to define a baseline (could be wrong, I didn't read the paper). He should have generated a few packages with a completely innocent name (and maybe some packages with just a GUID as a name) to see how much downloads / installs they get too.