number of libraries * number of versions ~= bits of information to classify you A good intuition is that although there's low certainty which site you have vis

number of libraries * number of versions ~= bits of information to classify you A good intuition is that although there's low certainty which site you have visited, there's high certainty which sites you HAVE NOT visited I think is easier to see how you can fingerprint someone based on the set of sites they HAVE NOT visited Edit: I'm not so sure anymore, I think that requires to test looots of libraries, it's not practical unless you are a really nasty ad company that tests hundreds of libraries in the background really sucking up your bandwidth