From: Sergei on
When talking about an indistinguishable encryption, it is required
that the probabilities of distinguishing two encrypted plaintext
differ negligibly. In contrast, in the definition of the differential
privacy, it is required that the probabilities of getting the same
results if computing a function on two data sets differing by at most
one element differ by a small multiplicative factor exp^(x).

The question is: Why not using a negligible function in the case for
differential privacy as well? Is it because having a multiplicative
factor allows to make differential privacy achievable when perturbing
the data using gaussian, binomial or Laplace distribution? Or are
there some fundamental reasons for it?

Sergei