Linear and Range Counting under Metric-based Local Differential Privacy

被引:0
作者
Xiang, Zhuolun [1 ,2 ]
Ding, Bolin [2 ]
He, Xi [3 ]
Zhou, Jingren [2 ]
机构
[1] UIUC, Urbana, IL 61801 USA
[2] Alibaba Grp, Hangzhou, Peoples R China
[3] Univ Waterloo, Waterloo, ON, Canada
来源
2020 IEEE INTERNATIONAL SYMPOSIUM ON INFORMATION THEORY (ISIT) | 2020年
关键词
D O I
10.1109/isit44484.2020.9173952
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Local differential privacy (LDP) enables private data sharing and analytics without the need for a trusted data collector. Error-optimal primitives (for, e.g., estimating means and item frequencies) under LDP have been well studied. For analytical tasks such as range queries, however, the best known error bound is dependent on the domain size of private data, which is potentially prohibitive. This deficiency is inherent as LDP protects the same level of indistinguishability between any pair of private data values for each data downer. In this paper, we utilize an extension of epsilon-LDP called Metric-LDP or E-LDP, where a metric E defines heterogeneous privacy guarantees for different pairs of private data values and thus provides a more flexible knob than epsilon does to relax LDP and tune utility-privacy trade-offs. We show that, under such privacy relaxations, for analytical workloads such as linear counting, multi-dimensional range counting queries, and quantile queries, we can achieve significant gains in utility. In particular, for range queries under E-LDP where the metric E is the L-1-distance function scaled by epsilon, we design mechanisms with errors independent on the domain sizes; instead, their errors depend on the metric E, which specifies in what granularity the private data is protected. We believe that the primitives we design for E-LDP will be useful in developing mechanisms for other analytical tasks, and encourage the adoption of LDP in practice. Full version of this paper at: https://arxiv.org/abs/1909.11778
引用
收藏
页码:908 / 913
页数:6
相关论文
共 38 条
  • [1] A. D. P. Team, 2017, APPLE MACHINE LEARNI
  • [2] Acharya J, 2019, 22 INT C ARTIFICIAL, V89
  • [3] Local Differential Privacy on Metric Spaces: optimizing the trade-off with utility
    Alvim, Mario S.
    Chatzikokolakis, Konstantinos
    Palamidessi, Catuscia
    Pazii, Anna
    [J]. IEEE 31ST COMPUTER SECURITY FOUNDATIONS SYMPOSIUM (CSF 2018), 2018, : 262 - 267
  • [4] Andres M. E., 2013, P 2013 ACM SIGSAC C, P901
  • [5] [Anonymous], 2017, NIPS
  • [6] Local, Private, Efficient Protocols for Succinct Histograms
    Bassily, Raef
    Smith, Adam
    [J]. STOC'15: PROCEEDINGS OF THE 2015 ACM SYMPOSIUM ON THEORY OF COMPUTING, 2015, : 127 - 135
  • [7] Heavy Hitters and the Structure of Local Privacy
    Bun, Mark
    Nelson, Jelani
    Stemmer, Uri
    [J]. PODS'18: PROCEEDINGS OF THE 37TH ACM SIGMOD-SIGACT-SIGAI SYMPOSIUM ON PRINCIPLES OF DATABASE SYSTEMS, 2018, : 435 - 447
  • [8] Chan THH, 2012, LECT NOTES COMPUT SC, V7501, P277, DOI 10.1007/978-3-642-33090-2_25
  • [9] Chatzikokolakis Konstantinos, 2013, Privacy Enhancing Technologies.13th International Symposium, PETS 2013. Proceedings: LNCS 7981, P82, DOI 10.1007/978-3-642-39077-7_5
  • [10] Chaudhuri K, 2019, ADV NEUR IN, V32