Privacy-Preserving Student Learning with Differentially Private Data-Free Distillation

被引:2
|
作者
Liu, Bochao [1 ,2 ]
Lu, Jianghu [1 ,2 ]
Wang, Pengju [1 ,2 ]
Zhang, Junjie [3 ]
Zeng, Dan [3 ]
Qian, Zhenxing [4 ]
Ge, Shiming [1 ,2 ]
机构
[1] Chinese Acad Sci, Inst Informat Engn, Beijing 100095, Peoples R China
[2] Univ Chinese Acad Sci, Sch Cyber Secur, Beijing 100049, Peoples R China
[3] Shanghai Univ, Sch Commun & Informat Engn, Shanghai 200444, Peoples R China
[4] Fudan Univ, Sch Comp Sci, Shanghai 200433, Peoples R China
基金
北京市自然科学基金;
关键词
differential privacy; teacher-student learning; knowledge distillation;
D O I
10.1109/MMSP55362.2022.9950001
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
Deep learning models can achieve high inference accuracy by extracting rich knowledge from massive well-annotated data, but may pose the risk of data privacy leakage in practical deployment. In this paper, we present an effective teacher-student learning approach to train privacy-preserving deep learning models via differentially private data-free distillation. The main idea is generating synthetic data to learn a student that can mimic the ability of a teacher well-trained on private data. In the approach, a generator is first pretrained in a data-free manner by incorporating the teacher as a fixed discriminator. With the generator, massive synthetic data can be generated for model training without exposing data privacy. Then, the synthetic data is fed into the teacher to generate private labels. Towards this end, we propose a label differential privacy algorithm termed selective randomized response to protect the label information. Finally, a student is trained on the synthetic data with the supervision of private labels. In this way, both data privacy and label privacy are well protected in a unified framework, leading to privacy-preserving models. Extensive experiments and analysis clearly demonstrate the effectiveness of our approach.
引用
收藏
页数:6
相关论文
共 50 条
  • [41] Privacy-Preserving and Approximately Truthful Local Electricity Markets: A Differentially Private VCG Mechanism
    Hoseinpour, Milad
    Hoseinpour, Mohammad
    Haghifam, Mahdi
    Haghifam, Mahmoud-Reza
    IEEE TRANSACTIONS ON SMART GRID, 2024, 15 (02) : 1991 - 2003
  • [42] SynTF: Synthetic and Differentially Private Term Frequency Vectors for Privacy-Preserving Text Mining
    Weggenmann, Benjamin
    Kerschbaum, Florian
    ACM/SIGIR PROCEEDINGS 2018, 2018, : 305 - 314
  • [43] Privacy-Preserving Classifier Learning
    Brickell, Justin
    Shmatikov, Vitaly
    FINANCIAL CRYPTOGRAPHY AND DATA SECURITY, 2009, 5628 : 128 - 147
  • [44] Privacy-Preserving Deep Learning
    Shokri, Reza
    Shmatikov, Vitaly
    CCS'15: PROCEEDINGS OF THE 22ND ACM SIGSAC CONFERENCE ON COMPUTER AND COMMUNICATIONS SECURITY, 2015, : 1310 - 1321
  • [45] Privacy-Preserving Deep Learning
    Shokri, Reza
    Shmatikov, Vitaly
    2015 53RD ANNUAL ALLERTON CONFERENCE ON COMMUNICATION, CONTROL, AND COMPUTING (ALLERTON), 2015, : 909 - 910
  • [46] Privacy-Preserving Machine Learning
    Chow, Sherman S. M.
    FRONTIERS IN CYBER SECURITY, 2018, 879 : 3 - 6
  • [47] Privacy-preserving data mining
    Agrawal, R
    Srikant, R
    SIGMOD RECORD, 2000, 29 (02) : 439 - 450
  • [48] Privacy-Preserving Data Publishing
    Liu, Ruilin
    Wang, Hui
    2010 IEEE 26TH INTERNATIONAL CONFERENCE ON DATA ENGINEERING WORKSHOPS (ICDE 2010), 2010, : 305 - 308
  • [49] Privacy-preserving data imputation
    Jagannathan, Geetha
    Wright, Rebecca N.
    ICDM 2006: SIXTH IEEE INTERNATIONAL CONFERENCE ON DATA MINING, WORKSHOPS, 2006, : 535 - +
  • [50] Privacy-Preserving Data Publishing
    Chen, Bee-Chung
    Kifer, Daniel
    LeFevre, Kristen
    Machanavajjhala, Ashwin
    FOUNDATIONS AND TRENDS IN DATABASES, 2009, 2 (1-2): : 1 - 167