Automatic Tuning of Privacy Budgets in Input-Discriminative Local Differential Privacy

被引:4
作者
Murakami, Takao [1 ]
Sei, Yuichi [2 ]
机构
[1] Natl Inst Adv Ind Sci & Technol, Cyber Phys Secur Res Ctr, Tokyo 1350064, Japan
[2] Univ Electrocommun, Grad Sch Informat & Engn, Tokyo 1828585, Japan
基金
日本学术振兴会;
关键词
Privacy; Data privacy; Tuning; Internet of Things; Databases; Proposals; Differential privacy; Frequency estimation; Internet of Things (IoT); local differential privacy (LDP); privacy budget; reidentification; CONFIDENCE-INTERVALS; PROPORTION;
D O I
10.1109/JIOT.2023.3267082
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Local differential privacy (LDP) and its variants have been recently studied to analyze personal data collected from Internet of Things (IoT) devices while strongly protecting user privacy. In particular, a recent study proposes a general privacy notion called input-discriminative LDP (ID-LDP), which introduces a privacy budget for each input value to deal with different levels of sensitivity. However, it is unclear how to set an appropriate privacy budget for each input value, especially, in current situations where reidentification is considered a major risk, e.g., in GDPR. Moreover, the possible number of input values can be very large in IoT. Consequently, it is also extremely difficult to manually check whether a privacy budget for each input value is appropriate. In this article, we propose algorithms to automatically tune privacy budgets in ID-LDP so that obfuscated data strongly prevent reidentification. We also propose a new instance of ID-LDP called one-budget ID-LDP (OneID-LDP) to prevent reidentification with high utility. Through comprehensive experiments using four real data sets, we show that existing instances of ID-LDP lack either utility or privacy-they overprotect personal data or are vulnerable to reidentification attacks. Then, we show that our OneID-LDP mechanisms with our privacy budget tuning algorithm provide much higher utility than LDP mechanisms while strongly preventing reidentification.
引用
收藏
页码:15990 / 16005
页数:16
相关论文
共 64 条
  • [21] Fanti Giulia, 2016, Proceedings on Privacy Enhancing Technologies, V2016, P41, DOI 10.1515/popets-2016-0015
  • [22] Feldman V, 2021, Arxiv, DOI arXiv:2012.12803
  • [23] Gehrke J, 2012, LECT NOTES COMPUT SC, V7417, P479
  • [24] Providing Input-Discriminative Protection for Local Differential Privacy
    Gu, Xiaolan
    Li, Ming
    Xiong, Li
    Cao, Yang
    [J]. 2020 IEEE 36TH INTERNATIONAL CONFERENCE ON DATA ENGINEERING (ICDE 2020), 2020, : 505 - 516
  • [25] Hartung J., 2008, Statistical Meta-Analysis With Applications
  • [26] A Discussion of Privacy Challenges in User Profiling with Big Data Techniques: The EEXCESS Use Case
    Hasan, Omar
    Habegger, Benjamin
    Brunie, Lionel
    Bennani, Nadia
    Damiani, Ernesto
    [J]. 2013 IEEE INTERNATIONAL CONGRESS ON BIG DATA, 2013, : 25 - 30
  • [27] Hastie T, 2009, ELEMENTS STAT LEARNI
  • [28] A Versatile Clustering Method for Electricity Consumption Pattern Analysis in Households
    Hino, Hideitsu
    Shen, Haoyang
    Murata, Noboru
    Wakao, Shinji
    Hayashi, Yasuhiro
    [J]. IEEE TRANSACTIONS ON SMART GRID, 2013, 4 (02) : 1048 - 1057
  • [29] Hull B., 2006, P 4 INT C EMB NETW S, P125, DOI DOI 10.1145/1182807.1182821
  • [30] Imola J, 2021, PROCEEDINGS OF THE 30TH USENIX SECURITY SYMPOSIUM, P983