Haven't read TFA yet, but at first glance this sounds similar to the approach used by the "Privacy Badger" browser extension - if it sees the same tracker on multiple sites, it "learns" and begins blocking it. Would it also be susceptible to similar information leaks with this threat model?