Given a collection of real data, we want to do some statistical analysis.
For example, we take sum of them and reveal it to someone else.
But, we want to guarantee some sort of privacy of individual data.
In particular, suppose we have $(x_1, \ldots,x_n) \in \mathbb{R}^n$ and $a = \sum_i^n x_i$ is revealed to someone (not the individual value $x_i$). How can we measure or bound the leakage of each value $_$ from $$?
Does differential privacy help?