Fedlabx: a practical and privacy-preserving framework for federated learning

被引:5
作者
Yan, Yuping [1 ,2 ]
Alshawki, Mohammed B. [1 ,4 ,5 ]
Zoltay, Marcell [2 ]
Gal, Marcell [2 ]
Hollos, Roland [2 ]
Jin, Yaochu [3 ]
Peter, Ligeti [1 ]
Tenyi, Akos [2 ]
机构
[1] Eotvos Lorand Univ, Dept Comp Algebra, Budapest, Hungary
[2] E Grp ICT Software Zrt, Smart Data Grp, Budapest, Hungary
[3] Bielefeld Univ, Fac Technol, D-33619 Bielefeld, Germany
[4] Furtwangen Univ, Inst Data Sci Cloud Comp & IT Secur, IDACUS, Furtwangen, Germany
[5] Univ Kufa, Dept Comp Sci, Najaf, Iraq
关键词
Federated learning; Kafka; Secure aggregation; Differential privacy;
D O I
10.1007/s40747-023-01184-3
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Federated learning (FL) draws attention in academia and industry due to its privacy-preserving capability in training machine learning models. However, there are still some critical security attacks and vulnerabilities, including gradients leakage and interference attacks. Meanwhile, communication is another bottleneck in basic FL schemes since large-scale FL parameter transmission leads to inefficient communication, latency, and slower learning processes. To overcome these shortcomings, different communication efficiency strategies and privacy-preserving cryptographic techniques have been proposed. However, a single method can only partially resist privacy attacks. This paper presents a practical, privacy-preserving scheme combining cryptographic techniques and communication networking solutions. We implement Kafka for message distribution, the Diffie-Hellman scheme for secure server aggregation, and gradient differential privacy for interference attack prevention. The proposed approach maintains training efficiency while being able to addressing gradients leakage problems and interference attacks. Meanwhile, the implementation of Kafka and Zookeeper provides asynchronous communication and anonymous authenticated computation with role-based access controls. Finally, we prove the privacy-preserving properties of the proposed solution via security analysis and empirically demonstrate its efficiency and practicality.
引用
收藏
页码:677 / 690
页数:14
相关论文
共 41 条
[1]  
Abadi M, 2016, PROCEEDINGS OF OSDI'16: 12TH USENIX SYMPOSIUM ON OPERATING SYSTEMS DESIGN AND IMPLEMENTATION, P265
[2]   Deep Learning with Differential Privacy [J].
Abadi, Martin ;
Chu, Andy ;
Goodfellow, Ian ;
McMahan, H. Brendan ;
Mironov, Ilya ;
Talwar, Kunal ;
Zhang, Li .
CCS'16: PROCEEDINGS OF THE 2016 ACM SIGSAC CONFERENCE ON COMPUTER AND COMMUNICATIONS SECURITY, 2016, :308-318
[3]   HIPAA regulations - A new era of medical-record privacy? [J].
Annas, GJ .
NEW ENGLAND JOURNAL OF MEDICINE, 2003, 348 (15) :1486-1490
[4]   Local Differential Privacy for Deep Learning [J].
Arachchige, Pathum Chamikara Mahawaga ;
Bertok, Peter ;
Khalil, Ibrahim ;
Liu, Dongxi ;
Camtepe, Seyit ;
Atiquzzaman, Mohammed .
IEEE INTERNET OF THINGS JOURNAL, 2020, 7 (07) :5827-5842
[5]  
Beutel DJ, 2022, Arxiv, DOI arXiv:2007.14390
[6]   Practical Secure Aggregation for Privacy-Preserving Machine Learning [J].
Bonawitz, Keith ;
Ivanov, Vladimir ;
Kreuter, Ben ;
Marcedone, Antonio ;
McMahan, H. Brendan ;
Patel, Sarvar ;
Ramage, Daniel ;
Segal, Aaron ;
Seth, Karn .
CCS'17: PROCEEDINGS OF THE 2017 ACM SIGSAC CONFERENCE ON COMPUTER AND COMMUNICATIONS SECURITY, 2017, :1175-1191
[7]  
Brendan McMahan H., CORR
[8]  
BUCHMANN J.A., 2013, Introduction to public key infrastructures
[9]  
Chen JJ, 2021, PACIFIC SYMPOSIUM ON BICOMPUTING 2021, P26
[10]   NEW DIRECTIONS IN CRYPTOGRAPHY [J].
DIFFIE, W ;
HELLMAN, ME .
IEEE TRANSACTIONS ON INFORMATION THEORY, 1976, 22 (06) :644-654