Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

> You can not completely anonymize data with any reliability.

Well... there's actually a field for that. I forgot what they call that field because of how niche it is but my friend at google is doing just that.

He said there are math theorem to prove that it's sufficiently anonymize.

He gave an example of how Netflix competition with the data they gave researchers were able to deanonymize it. And his job was to prevent that at google.

I can see why if you're trying to sell users data while maintaining privacy.



Mozilla currently uses Google Analytics for tracking, with "IP Anonymization" enabled.

Which, according to Google’s FAQ, https://support.google.com/analytics/answer/2763052?hl=en, just blanks out the last byte of the IP.

Which is useless, because it still includes enough personalized data as to be completely and utterly reversible.


Not to mention, if we're talking about IPv6 addresses, they often get allocated as /64 (or even /60 or /56) blocks. Such "anonymization" becomes useless at that point.


Google Analytics has nothing to do with this. As clearly linked in the mailing list, you can read the paper and source code for the client-side differential privacy tech used.



The linked Wikipedia page actually mentions that competition: https://en.wikipedia.org/wiki/Differential_privacy#Netflix_P...




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: