Partial Privacy Data Analysis

[jinmingjian]

it is interesting that if we provide some foundations to enable partial privacy in the bigdata analysis, considering the FHE is too expensive.

[alexey-milovidov]

There was a prototype implementation for differential privacy in ClickHouse, but it is not very practical.

[jinmingjian]

@alexey-milovidov Great! "Differential privacy", as my understanding, will introduce lossy statistical perturbation. This is great for the privacy for the users of apps on the top of data system. My viewpoint of this issue is the privacy for the users of data system itself. So FHE listed. Of course, they are orthogonal in some extent.