- From: Joseph Lorenzo Hall <joe@cdt.org>
- Date: Fri, 18 Sep 2015 16:02:13 -0400
- To: David Singer <singer@apple.com>
- Cc: Christine Runnegar <runnegar@isoc.org>, "public-privacy (W3C mailing list)" <public-privacy@w3.org>
On Fri, Sep 18, 2015 at 3:37 PM, David Singer <singer@apple.com> wrote: > >>> • Does the data record contain elements that would enable re-correlation when combined with other datasets through the property of intersection? >>> • No (just audio/video) >>> This seems like a hard question... on the one hand, if a "face" is enough from which to derive a facial pattern that you can correlate with other databases of facial patterns, then the answer would seem to be yes (although I don't know of any public databases of facial biometrics). Maybe there's a better way to get at what this question wants to get at? Does anyone remember what the impetus for this question is? or can we think of examples in a spec that we'd definitely want to catch with this question? > > Yes. Isn’t this getting at the problem that if I know someone’s gender, birthday, zip-code and one other datum (I forget what), I can almost certainly identify them, even though any one of these looks innocuous? The US-based privacy law academic community would refer to this as "the mosiac theory" (that any one datum is not revealing but in concert more than one can be quite revealing). Sweeney [1] published the first analysis that revealed the power of the gender, birthday, and zip code three-tuple using 1990 US census data, although Golle [2] updated that and found that in ten years it had become less uniquely identifying (presumably because of increased urban concentration of the US population?). I don't think anyone has repeated the analysis with 2010 US census data, which would be cool. But I digress! I'm having a hard time thinking of ways to make this particular question easier for a spec-author to handle (and for us to evaluate) without potentially loosing important privacy thinking we or they should do in the process. Hmmmm... best, Joe [1]: http://dataprivacylab.org/projects/identifiability/paper1.pdf [2]: https://crypto.stanford.edu/~pgolle/papers/census.pdf -- Joseph Lorenzo Hall Chief Technologist Center for Democracy & Technology 1634 I ST NW STE 1100 Washington DC 20006-4011 (p) 202-407-8825 (f) 202-637-0968 joe@cdt.org PGP: https://josephhall.org/gpg-key fingerprint: 3CA2 8D7B 9F6D DBD3 4B10 1607 5F86 6987 40A9 A871
Received on Friday, 18 September 2015 20:03:02 UTC