Henry Reich
patreon
Collab with the US Census Bureau
🕑 Added 2019-09-12 13:04:54 +0000 UTCThe Chief Scientist of the US Census Bureau reached out to me and asked if I would help explain some of the math they're using in the 2020 Census to protect privacy. So I got to learn a bit about computer science and the mathematics of privacy.
Thanks as always for your support.
Henry
Comments
Neal McBurnett
I love seeing a video about epsilon-differential privacy! But I wish there were some specific quantitative examples e.g. of how much jittering they do for a few sample data points. And I must say the quote at 11:45from the chief scientist is what I'd want the audience to take away from this. I think you tended to confuse the issue by making it seem like you always need to jitter to achieve any epsilon given any data release. But surely I can give you just the mean and median for the total US population without any need for jitter, or any significant risk to privacy. I.e. epsilon is only high enough to worry about when we give out lots of detailed statistics.
