English

Have it your way: Individualized Privacy Assignment for DP-SGD

Machine Learning 2024-01-22 v2 Artificial Intelligence Cryptography and Security

Abstract

When training a machine learning model with differential privacy, one sets a privacy budget. This budget represents a maximal privacy violation that any user is willing to face by contributing their data to the training set. We argue that this approach is limited because different users may have different privacy expectations. Thus, setting a uniform privacy budget across all points may be overly conservative for some users or, conversely, not sufficiently protective for others. In this paper, we capture these preferences through individualized privacy budgets. To demonstrate their practicality, we introduce a variant of Differentially Private Stochastic Gradient Descent (DP-SGD) which supports such individualized budgets. DP-SGD is the canonical approach to training models with differential privacy. We modify its data sampling and gradient noising mechanisms to arrive at our approach, which we call Individualized DP-SGD (IDP-SGD). Because IDP-SGD provides privacy guarantees tailored to the preferences of individual users and their data points, we find it empirically improves privacy-utility trade-offs.

Keywords

Cite

@article{arxiv.2303.17046,
  title  = {Have it your way: Individualized Privacy Assignment for DP-SGD},
  author = {Franziska Boenisch and Christopher Mühl and Adam Dziedzic and Roy Rinberg and Nicolas Papernot},
  journal= {arXiv preprint arXiv:2303.17046},
  year   = {2024}
}

Comments

Published at NeurIPS'2024

R2 v1 2026-06-28T09:40:43.135Z