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
  • [1] Distributed Linear Bandits With Differential Privacy
    Li, Fengjiao
    Zhou, Xingyu
    Ji, Bo
    IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING, 2024, 11 (03): : 3161 - 3173
  • [2] Global and Local Differential Privacy for Collaborative Bandits
    Wang, Huazheng
    Zhao, Qian
    Wu, Qingyun
    Chopra, Shubham
    Khaitan, Abhinav
    Wang, Hongning
    RECSYS 2020: 14TH ACM CONFERENCE ON RECOMMENDER SYSTEMS, 2020, : 150 - 159
  • [3] Optimal Learning Policies for Differential Privacy in Multi-armed Bandits
    Wang, Siwei
    Zhu, Jun
    JOURNAL OF MACHINE LEARNING RESEARCH, 2024, 25
  • [4] Privacy Amplification via Shuffling for Linear Contextual Bandits
    Garcelon, Evrard
    Chaudhuri, Kamalika
    Perchet, Vianney
    Pirotta, Matteo
    INTERNATIONAL CONFERENCE ON ALGORITHMIC LEARNING THEORY, VOL 167, 2022, 167
  • [5] Limiting Privacy Breaches in Differential Privacy
    Ouyang Jia
    Yin Jian
    Liu Shao-Peng
    PROCEEDINGS OF THE 3RD INTERNATIONAL CONFERENCE ON COMPUTER SCIENCE AND SERVICE SYSTEM (CSSS), 2014, 109 : 657 - 664
  • [6] Differential privacy and robust statistics in high dimensions
    Liu, Xiyang
    Kong, Weihao
    Oh, Sewoong
    CONFERENCE ON LEARNING THEORY, VOL 178, 2022, 178
  • [7] Privacy at Scale: Local Differential Privacy in Practice
    Cormode, Graham
    Jha, Somesh
    Kulkarni, Tejas
    Li, Ninghui
    Srivastava, Divesh
    Wang, Tianhao
    SIGMOD'18: PROCEEDINGS OF THE 2018 INTERNATIONAL CONFERENCE ON MANAGEMENT OF DATA, 2018, : 1655 - 1658
  • [8] Optimal Distribution of Privacy Budget in Differential Privacy
    Bkakria, Anis
    Tasidou, Aimilia
    Cuppens-Boulahia, Nora
    Cuppens, Frederic
    Bouattour, Fatma
    Ben Fredj, Feten
    RISKS AND SECURITY OF INTERNET AND SYSTEMS, 2019, 11391 : 222 - 236
  • [9] Differential privacy in practice
    Nguyen, Hiep H.
    Kim, Jong
    Kim, Yoonho
    Journal of Computing Science and Engineering, 2013, 7 (03) : 177 - 186
  • [10] Differential privacy in deep learning: Privacy and beyond
    Wang, Yanling
    Wang, Qian
    Zhao, Lingchen
    Wang, Cong
    FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE, 2023, 148 : 408 - 424