14
52

Learning discrete distributions: user vs item-level privacy

Abstract

Much of the literature on differential privacy focuses on item-level privacy, where loosely speaking, the goal is to provide privacy per item or training example. However, recently many practical applications such as federated learning require preserving privacy for all items of a single user, which is much harder to achieve. Therefore understanding the theoretical limit of user-level privacy becomes crucial. We study the fundamental problem of learning discrete distributions over kk symbols with user-level differential privacy. If each user has mm samples, we show that straightforward applications of Laplace or Gaussian mechanisms require the number of users to be O(k/(mα2)+k/ϵα)\mathcal{O}(k/(m\alpha^2) + k/\epsilon\alpha) to achieve an 1\ell_1 distance of α\alpha between the true and estimated distributions, with the privacy-induced penalty k/ϵαk/\epsilon\alpha independent of the number of samples per user mm. Moreover, we show that any mechanism that only operates on the final aggregate counts should require a user complexity of the same order. We then propose a mechanism such that the number of users scales as O~(k/(mα2)+k/mϵα)\tilde{\mathcal{O}}(k/(m\alpha^2) + k/\sqrt{m}\epsilon\alpha) and hence the privacy penalty is Θ~(m)\tilde{\Theta}(\sqrt{m}) times smaller compared to the standard mechanisms in certain settings of interest. We further show that the proposed mechanism is nearly-optimal under certain regimes. We also propose general techniques for obtaining lower bounds on restricted differentially private estimators and a lower bound on the total variation between binomial distributions, both of which might be of independent interest.

View on arXiv
Comments on this paper