Towards Benchmarking Privacy Risk for Differential Privacy: A Survey

被引:0
作者
Prokhorenkov, Dmitry [1 ]
Cao, Yang [2 ]
机构
[1] Tech Univ Munich, Garching, Germany
[2] Hokkaido Univ, Sapporo, Hokkaido, Japan
来源
PROCEEDINGS OF THE 10TH ACM INTERNATIONAL CONFERENCE ON SYSTEMS FOR ENERGY-EFFICIENT BUILDINGS, CITIES, AND TRANSPORTATION, BUILDSYS 2023 | 2023年
关键词
data privacy; GDPR; differential privacy; privacy risk; attack; THREATS;
D O I
10.1145/3600100.3625373
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
While utilizing the Differential Privacy (DP) method, there are still many open issues regarding the method's effectiveness and reliability, including multiple difficulties in measuring the privacy level, which leads to implications in assessing risks while operating with personal data. Today, many experts and researchers emphasize the need and importance of utilizing DP for various kinds of data and tasks. However, to the best of our knowledge, the studies have yet to analyze the DP method regarding privacy risk assessment, thereby assessing the risks while utilizing DP and in the context of known types of privacy attacks. As a result, this study examines the existing privacy risks for various DP types in the context of present types of attacks. Also, the concept of privacy risk for DP will be analyzed, along with the corresponding metrics and their utility measurement. Our research method relies on a literature review; as a result, studies published from 2010 to 2023 were reviewed. Selected articles we examined based on the existing types of attacks, methods and metrics used to assess privacy risks. Based on this, we advanced the concept of privacy risk since none of the scientific studies clearly established the notation privacy risk for DP. Thus, we seek to explain the concept of privacy risk for DP in the context of existing types of attacks, thereby enabling DP utilization concerning the GDPR.
引用
收藏
页码:322 / 327
页数:6
相关论文
共 77 条
  • [1] Altman M., 2021, Boston University Journal of Science & Technology Law, V27, P1
  • [2] Amin K, 2019, PR MACH LEARN RES, V97
  • [3] [Anonymous], 2011, Privacy enhancing technologies: A review
  • [4] Bagdasaryan E, 2019, ADV NEUR IN, V32
  • [5] Balston George, 2022, Implications of Emerging Privacy Enhancing Technologies for UK Surveillance Policy
  • [6] Banerjee Mishtu, 2011, Secure Data Management. Proceedings 8th VLDB Workshop (SDM2011), P1, DOI 10.1007/978-3-642-23556-6_1
  • [7] Bastos Daniel, 2018, P 4 ANN IOT SEC FDN, V4, P1
  • [8] Assessing Differentially Private Variational Autoencoders Under Membership Inference
    Bernau, Daniel
    Robl, Jonas
    Kerschbaum, Florian
    [J]. DATA AND APPLICATIONS SECURITY AND PRIVACY XXXVI, DBSEC 2022, 2022, 13383 : 3 - 14
  • [9] Bernau D, 2020, Arxiv, DOI arXiv:1912.11328
  • [10] Comparing Local and Central Differential Privacy Using Membership Inference Attacks
    Bernau, Daniel
    Robl, Jonas
    Grassal, Philip W.
    Schneider, Steffen
    Kerschbaum, Florian
    [J]. DATA AND APPLICATIONS SECURITY AND PRIVACY XXXV, 2021, 12840 : 22 - 42