Improving the Privacy and Practicality of Objective Perturbation for Differentially Private Linear Learners
Improving the Privacy and Practicality of Objective Perturbation for Differentially Private Linear Learners
复制标题
DOI：
10.48550/arxiv.2401.00583
复制
发表时间：
2023-12
期刊：
影响因子：
--
通讯作者：
Rachel Redberg;Antti Koskela;Yu-Xiang Wang
中科院分区：
文献类型：
--
作者：
Rachel Redberg;Antti Koskela;Yu-Xiang Wang
在隐私保护机器学习的竞技场中，差分隐私随机梯度下降（DP-SGD）在受欢迎程度和兴趣方面超过了目标扰动机制。虽然DP-SGD在通用性方面无与伦比，但它需要一个非平凡的隐私开销（用于私下调整模型的超参数）和计算复杂性，这对于线性和逻辑回归等简单模型来说可能是奢侈的。本文改进了目标扰动机制，更严格的隐私分析和新的计算工具，使其在无约束凸广义线性问题上与DP-SGD竞争。
In the arena of privacy-preserving machine learning, differentially private stochastic gradient descent (DP-SGD) has outstripped the objective perturbation mechanism in popularity and interest. Though unrivaled in versatility, DP-SGD requires a non-trivial privacy overhead (for privately tuning the model's hyperparameters) and a computational complexity which might be extravagant for simple models such as linear and logistic regression. This paper revamps the objective perturbation mechanism with tighter privacy analyses and new computational tools that boost it to perform competitively with DP-SGD on unconstrained convex generalized linear problems.