The Cost of Privacy in Generalized Linear Models: Algorithms and Minimax Lower Bounds

The Cost of Privacy in Generalized Linear Models: Algorithms and Minimax Lower Bounds
复制标题

广义线性模型中的隐私成本:算法和极小极大下界

DOI:
10.48550/arxiv.2205.03014
复制
发表时间:
2020
期刊:
ArXiv
影响因子:
--
通讯作者:
Linjun Zhang
Linjun Zhang
中科院分区:
--
文献类型:
--
作者:
T. Cai;Yichen Wang;Linjun Zhang

文献摘要

参考文献

被引文献

相似文献

我们通过构造投影梯度下降的私有版本,提出了用于低维和高维稀疏广义线性模型(GLMs)参数估计的差分私有算法。我们通过描述其统计性能和建立隐私约束的最小极大下界来证明所提出的算法几乎是速率最优的。通过一种基于Stein引理的新技术,推广了隐私约束下界的跟踪攻击技术,得到了下界。这个下界的论点可以是独立的兴趣,因为它适用于一般的参数模型。通过模拟和实际数据实验验证了算法的数值性能。
We propose differentially private algorithms for parameter estimation in both low-dimensional and high-dimensional sparse generalized linear models (GLMs) by constructing private versions of projected gradient descent. We show that the proposed algorithms are nearly rate-optimal by characterizing their statistical performance and establishing privacy-constrained minimax lower bounds for GLMs. The lower bounds are obtained via a novel technique, which is based on Stein's Lemma and generalizes the tracing attack technique for privacy-constrained lower bounds. This lower bound argument can be of independent interest as it is applicable to general parametric models. Simulated and real data experiments are conducted to demonstrate the numerical performance of our algorithms.
DOI: 10.1016/s0092-8240(89)80052-7
发表时间: 1989-01-01
影响因子: 3.5
作者:
ARRATIA, R;GORDON, L
通讯作者: GORDON, L
DOI: --
发表时间: 2018-05
期刊: --
影响因子: --
作者:
Gautam Kamath;Jerry Li;Vikrant Singhal;Jonathan Ullman
通讯作者: Gautam Kamath;Jerry Li;Vikrant Singhal;Jonathan Ullman
DOI: 10.1007/s00440-018-00896-9
发表时间: 2017-06
影响因子: 2
作者:
P. Sur;Yuxin Chen;E. Candès
通讯作者: P. Sur;Yuxin Chen;E. Candès
DOI: 10.1016/j.ekir.2018.04.002
发表时间: 2018-04-16
影响因子: 6
作者:
Kume S;Nagasu H;Nangaku M;Nishiyama A;Nakamoto H;Kashihara N
通讯作者: Kashihara N
DOI: 10.1002/adts.202200006
发表时间: 2022-03-31
影响因子: 3.3
作者:
Huang, Guo Shuai;Li, Si Jia;Cao, Xiang Yu
通讯作者: Cao, Xiang Yu