arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.01607math.STcs.ITcs.LGmath.ITstat.MLstat.TH
收藏

最小最大最优差分隐私合成数据用于平滑查询

Minimax optimal differentially private synthetic data for smooth queries

Department of Mathematics, University of Southern California(南加州大学数学系);Department of Mathematics, University of California San Diego(加州圣地亚哥大学数学系)

Rundong Ding, Yiyun He, Yizhe Zhu

AI总结:

本文研究了如何生成具有(ε,δ)差分隐私的合成数据,以在保证个体隐私的同时,为有意义的下游分析提供强效用保证。提出了一种多项式时间算法,实现了最小最大误差率O_{k,d}(n^{-min{1, k/d}}),并建立了针对k-平滑查询的首个最小最大下界。

摘要:

差分隐私合成数据使敏感数据集的共享和分析成为可能,同时为个体贡献者提供严格的隐私保证。一个核心挑战是为有意义的下游分析提供强效用保证。许多现有方法确保在广泛的查询类上具有均匀的准确性,如所有Lipschitz函数,但这种通用性往往导致对实际感兴趣的统计量的次优速率。由于许多常见数据分析查询的平滑性超出了最坏情况Lipschitz界所捕捉的范围,我们询问是否可以利用这种额外的结构来提高效用。我们研究了从大小为n的数据集生成(ε,δ)差分隐私合成数据的问题,该数据集支持在超立方体[-1,1]^d上,具有对所有具有受界导数的平滑查询的均匀效用保证。我们提出了一种多项式时间算法,实现了最小最大误差率O_{k,d}(n^{-min{1, k/d}}),除了一个log(n)因子。这一特征揭示了k=d处的相变。我们的结果推广了Chebyshev矩匹配框架(Musco等,2025;Wang等,2016),并且严格改进了在\citep{wang2016differentially}中为k-平滑查询建立的误差率。此外,我们建立了针对k-平滑查询的首个最小最大下界,扩展了Boedihardjo等(2024)中关于ε-差分隐私的Wasserstein下界。

英文摘要:

Differentially private synthetic data enables the sharing and analysis of sensitive datasets while providing rigorous privacy guarantees for individual contributors. A central challenge is to achieve strong utility guarantees for meaningful downstream analysis. Many existing methods ensure uniform accuracy over broad query classes, such as all Lipschitz functions, but this level of generality often leads to suboptimal rates for statistics of practical interest. Since many common data analysis queries exhibit smoothness beyond what worst-case Lipschitz bounds capture, we ask whether exploiting this additional structure can yield improved utility. We study the problem of generating $(\varepsilon,δ)$-differentially private synthetic data from a dataset of size $n$ supported on the hypercube $[-1,1]^d$, with utility guarantees uniformly for all smooth queries having bounded derivatives up to order $k$. We propose a polynomial-time algorithm that achieves a minimax error rate of $O_{k,d}(n^{-\min \{1, \frac{k}{d}\}})$, up to a $\log(n)$ factor. This characterization uncovers a phase transition at $k=d$. Our results generalize the Chebyshev moment matching framework of (Musco et al., 2025; Wang et al., 2016) and strictly improve the error rates for $k$-smooth queries established in \citep{wang2016differentially}. Moreover, we establish the first minimax lower bound for the utility of $(\varepsilon,δ)$-differentially private synthetic data with respect to $k$-smooth queries, extending the Wasserstein lower bound for $\varepsilon$-differential privacy in (Boedihardjo et al., 2024).

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑