Deanonymization that works well relies on: (i) correlating social graphs (ii) correlating likes/dislikes/reviews etc. across different networks. (iii) Lots of

Deanonymization that works well relies on: (i) correlating social graphs (ii) correlating likes/dislikes/reviews etc. across different networks. (iii) Lots of data to do (i) and (ii) And it's still difficult to do for random people on the internet (as opposed to the NSA or serious attackers such as those willing to put in the effort to crawl and analyze the entire linkedin graph.) I believe deanonymization based on just textual analysis is still a little bit of an academic effort. Anonymity does buy quite a bit - especially on a forum like HN - where there isn't a social graph and the like/dislike information is private.