Concentrated Differential Privacy for Bandits

被引:0
|
作者
Azize, Achraf [1 ]
Basu, Debabrota [1 ]
机构
[1] Univ Lille, Inria, CNRS, Cent Lille,CRIStAL,Equipe Scool, Lille, France
来源
IEEE CONFERENCE ON SAFE AND TRUSTWORTHY MACHINE LEARNING, SATML 2024 | 2024年
关键词
Differential Privacy; Multi-armed Bandits; Regret Analysis; Lower bounds;
D O I
10.1109/SaTML59370.2024.00013
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Bandits serve as the theoretical foundation of sequential learning and an algorithmic foundation of modern recommender systems. However, recommender systems often rely on user-sensitive data, making privacy a critical concern. This paper contributes to the understanding of Differential Privacy (DP) in bandits with a trusted centralised decision-maker, and especially the implications of ensuring zero Concentrated Differential Privacy (zCDP). First, we formalise and compare different adaptations of DP to bandits, depending on the considered input and the interaction protocol. Then, we propose three private algorithms, namely AdaC-UCB, AdaC-GOPE and AdaC-OFUL, for three bandit settings, namely finite-armed bandits, linear bandits, and linear contextual bandits. The three algorithms share a generic algorithmic blueprint, i.e. the Gaussian mechanism and adaptive episodes, to ensure a good privacy-utility trade-off. We analyse and upper bound the regret of these three algorithms. Our analysis shows that in all of these settings, the prices of imposing zCDP are (asymptotically) negligible in comparison with the regrets incurred oblivious to privacy. Next, we complement our regret upper bounds with the first minimax lower bounds on the regret of bandits with zCDP. To prove the lower bounds, we elaborate a new proof technique based on couplings and optimal transport. We conclude by experimentally validating our theoretical results for the three different settings of bandits.
引用
收藏
页码:78 / 109
页数:32
相关论文
共 50 条
  • [41] Randomized Response and Differential Privacy
    Ioannidis, Andreas
    Litke, Antonios
    Papadakis, Nikolaos
    17TH ACM INTERNATIONAL CONFERENCE ON PERVASIVE TECHNOLOGIES RELATED TO ASSISTIVE ENVIRONMENTS, PETRA 2024, 2024, : 600 - 605
  • [42] Differential Privacy on Edge Computing
    Jiang, Xiyu
    Tsou, Yao-Tung
    Kuo, Sy-Yen
    IEEE NANOTECHNOLOGY MAGAZINE, 2023, 17 (06) : 14 - 22
  • [43] A framework for adaptive differential privacy
    Winograd-Cort D.
    Haeberlen A.
    Roth A.
    Pierce B.C.
    2017, Association for Computing Machinery (01)
  • [44] Lower Bounds in Differential Privacy
    De, Anindya
    THEORY OF CRYPTOGRAPHY (TCC 2012), 2012, 7194 : 321 - 338
  • [45] Differential privacy in probabilistic systems
    Yang, Jiannan
    Cao, Yongzhi
    Wang, Hanpin
    INFORMATION AND COMPUTATION, 2017, 254 : 84 - 104
  • [46] Investigating Differential Privacy Outcomes
    Hasuda, Davi Grossi
    Bezerra, Juliana de Melo
    ENTERPRISE INFORMATION SYSTEMS, ICEIS 2021, 2022, 455 : 388 - 400
  • [47] RELATIONAL *-LIFTINGS FOR DIFFERENTIAL PRIVACY
    Barthe, Gilles
    Espitau, Thomas
    Hsu, Justin
    Sato, Tetsuya
    Strub, Pierre-Yves
    LOGICAL METHODS IN COMPUTER SCIENCE, 2019, 15 (04) : 18:1 - 18:32
  • [48] Prescriptive analytics with differential privacy
    Haripriya Harikumar
    Santu Rana
    Sunil Gupta
    Thin Nguyen
    Ramachandra Kaimal
    Svetha Venkatesh
    International Journal of Data Science and Analytics, 2022, 13 : 123 - 138
  • [49] Landmark Privacy: Configurable Differential Privacy Protection for Time Series
    Katsomallos, Manos
    Tzompanaki, Katerina
    Kotzinos, Dimitris
    CODASPY'22: PROCEEDINGS OF THE TWELVETH ACM CONFERENCE ON DATA AND APPLICATION SECURITY AND PRIVACY, 2022, : 179 - 190
  • [50] Generating Perturbations with Hilbert Curves and Differential Privacy for Location Privacy
    Wang, Na
    Yu, Haiyang
    PROCEEDINGS OF THE 2017 INTERNATIONAL CONFERENCE ON MECHANICAL, ELECTRONIC, CONTROL AND AUTOMATION ENGINEERING (MECAE 2017), 2017, 61 : 95 - 100