Hacker News Reader

Comment by shaggyfrog | Hacker News Reader

I think it got 100% for me, on Safari/Mac.

Note that it doesn't need to be 100% accurate to be effective. If it guesses better than 50% (i.e. coin flip), then it could be used to give guesses with at least some confidence. No different than analyzing any other noisy dataset. Because this all works client-side, it can also be done quite invisibly.

View on HN

Right, but those applications will be for things like online advertising, which is already tracking your visits and/or just assuming you use the popular sites anyway. Can this be used to violate privacy in a meaningful manner if it misses 50% or more of the time? You'd have complete plausible deniability if you were pinned to have accessed some site you don't want people to know you accessed. It can't be used as evidence in anything. What would the threat look like?

I oppose user tracking not necessarily because there's something to hide, but because a sense privacy is a fundamental component of a sense of independence, and for me, independence is one of the critical components of happiness.

It would have to guess better than anonymous modelling, not 50%. I'd happily bet even odds layout that each of my site visitors visit Google.

If the question is "Has this user visited site X?" then I'd hope that any kind of modelling is better than 50%, as simulating a coin toss would be at least as good.

there is already way better and reliable methods to accomplish same goal. e.g

http://ajaxian.com/archives/spyjax-using-avisited-to-test-yo...

I was under the impression that attempts had been made to hide any effects of :visited styles on the accessible DOM to stop this from working. There's a particularly good article on Mozilla's attempts [1], and the relevant bug on Bugzilla. [2]

[1]: http://dbaron.org/mozilla/visited-privacy

[2]: https://bugzilla.mozilla.org/show_bug.cgi?id=147777

Reply on news.ycombinator.com