Locally Differentially Private Heavy Hitter Identification

被引:53
|
作者
Wang, Tianhao [1 ]
Li, Ninghui [1 ]
Jha, Somesh [2 ]
机构
[1] Purdue Univ, Dept Comp Sci, W Lafayette, IN 47907 USA
[2] Univ Wisconsin, Dept Comp Sci, Madison, WI 53706 USA
关键词
Protocols; Frequency estimation; Differential privacy; Frequency-domain analysis; Estimation; Privacy; Sociology; Local differential privacy; heavy hitter;
D O I
10.1109/TDSC.2019.2927695
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
The notion of Local Differential Privacy (LDP) enables users to answer sensitive questions while preserving their privacy. The basic LDP frequency oracle protocol enables the aggregator to estimate the frequency of any value. But when the domain of input values is large, finding the most frequent values, also known as the heavy hitters, by estimating the frequencies of all possible values, is computationally infeasible. In this paper, we propose an LDP protocol for identifying heavy hitters. In our proposed protocol, which we call Prefix Extending Method (PEM), users are divided into groups, with each group reporting a prefix of her value. We analyze how to choose optimal parameters for the protocol and identify two design principles for designing LDP protocols with high utility. Experiments show that under the same privacy guarantee and computational cost, PEM has better utility on both synthetic and real-world datasets than existing solutions.
引用
收藏
页码:982 / 993
页数:12
相关论文
共 50 条
  • [31] On the Utility Gain of Iterative Bayesian Update for Locally Differentially Private Mechanisms
    Arcolezi, Heber H.
    Cerna, Selene
    Palamidessi, Catuscia
    DATA AND APPLICATIONS SECURITY AND PRIVACY XXXVII, DBSEC 2023, 2023, 13942 : 165 - 183
  • [32] Differentially Private Filtering
    Le Ny, Jerome
    Pappas, George J.
    IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2014, 59 (02) : 341 - 354
  • [33] Locally Differentially Private Embedding Models in Distributed Fraud Prevention Systems
    Perez, Iker
    Wong, Jason
    Skalski, Piotr
    Burrell, Stuart
    Mortier, Richard
    McAuley, Derek
    Sutton, David
    2023 23RD IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS, ICDMW 2023, 2023, : 475 - 484
  • [34] Locally Differentially-Private Randomized Response for Discrete Distribution Learning
    Pastore, Adriano
    Gastpar, Michael
    JOURNAL OF MACHINE LEARNING RESEARCH, 2021, 22
  • [35] Differentially Private Publication of Vertically Partitioned Data
    Tang, Peng
    Cheng, Xiang
    Su, Sen
    Chen, Rui
    Shao, Huaxi
    IEEE TRANSACTIONS ON DEPENDABLE AND SECURE COMPUTING, 2021, 18 (02) : 780 - 795
  • [36] Locally Differentially Private Quantile Summary Aggregation in Wireless Sensor Networks
    Aseeri, Aishah
    Zhang, Rui
    INTELLIGENT INFORMATION AND DATABASE SYSTEMS, ACIIDS 2022, PT I, 2022, 13757 : 364 - 376
  • [37] Proving that Programs Are Differentially Private
    McIver, Annabelle
    Morgan, Carroll
    PROGRAMMING LANGUAGES AND SYSTEMS, APLAS 2019, 2019, 11893 : 3 - 18
  • [38] ON DIFFERENTIALLY PRIVATE KALMAN FILTERING
    Degue, Kwassi H.
    Le Ny, Jerome
    2017 IEEE GLOBAL CONFERENCE ON SIGNAL AND INFORMATION PROCESSING (GLOBALSIP 2017), 2017, : 487 - 491
  • [39] Differentially Private Distributed Sensing
    Fink, Glenn A.
    2016 IEEE 3RD WORLD FORUM ON INTERNET OF THINGS (WF-IOT), 2016, : 216 - 221
  • [40] Differentially Private LQ Control
    Yazdani, Kasra
    Jones, Austin
    Leahy, Kevin
    Hale, Matthew
    IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2023, 68 (02) : 1061 - 1068