CodedPaddedFL and CodedSecAgg: Straggler Mitigation and Secure Aggregation in Federated Learning

被引:24
作者
Schlegel, Reent [1 ,2 ]
Kumar, Siddhartha [1 ,3 ]
Rosnes, Eirik [1 ]
Graell i Amat, Alexandre [1 ,4 ]
机构
[1] Simula UiB, N-5006 Bergen, Norway
[2] OHB Digital Connect GmbH, D-28359 Bremen, Germany
[3] Qamcom Res & Technol, S-41285 Gothenburg, Sweden
[4] Chalmers Univ Technol, Dept Elect Engn, S-41296 Gothenburg, Sweden
基金
瑞典研究理事会;
关键词
Servers; Codes; Resilience; Data models; Training; Decoding; Data privacy; Coded distributed computing; federated learning; gradient codes; linear regression; privacy; secure aggregation; straggler mitigation; MODEL;
D O I
10.1109/TCOMM.2023.3244243
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
We present two novel federated learning (FL) schemes that mitigate the effect of straggling devices by introducing redundancy on the devices' data across the network. Compared to other schemes in the literature, which deal with stragglers or device dropouts by ignoring their contribution, the proposed schemes do not suffer from the client drift problem. The first scheme, CodedPaddedFL, mitigates the effect of stragglers while retaining the privacy level of conventional FL. It combines one-time padding for user data privacy with gradient codes to yield straggler resiliency. The second scheme, CodedSecAgg, provides straggler resiliency and robustness against model inversion attacks and is based on Shamir's secret sharing. We apply CodedPaddedFL and CodedSecAgg to a classification problem. For a scenario with 120 devices, CodedPaddedFL achieves a speed-up factor of 18 for an accuracy of 95% on the MNIST dataset compared to conventional FL. Furthermore, it yields similar performance in terms of latency compared to a recently proposed scheme by Prakash et al. without the shortcoming of additional leakage of private data. CodedSecAgg outperforms the state-of-the-art secure aggregation scheme LightSecAgg by a speed-up factor of 6.6-18.7 for the MNIST dataset for an accuracy of 95%.
引用
收藏
页码:2013 / 2027
页数:15
相关论文
共 52 条
  • [1] Secure Single-Server Aggregation with (Poly)Logarithmic Overhead
    Bell, James Henry
    Bonawitz, Kallista A.
    Gascon, Adria
    Lepoint, Tancrede
    Raykova, Mariana
    [J]. CCS '20: PROCEEDINGS OF THE 2020 ACM SIGSAC CONFERENCE ON COMPUTER AND COMMUNICATIONS SECURITY, 2020, : 1253 - 1269
  • [2] Bonawitz K, 2019, ARXIV
  • [3] Practical Secure Aggregation for Privacy-Preserving Machine Learning
    Bonawitz, Keith
    Ivanov, Vladimir
    Kreuter, Ben
    Marcedone, Antonio
    McMahan, H. Brendan
    Patel, Sarvar
    Ramage, Daniel
    Segal, Aaron
    Seth, Karn
    [J]. CCS'17: PROCEEDINGS OF THE 2017 ACM SIGSAC CONFERENCE ON COMPUTER AND COMMUNICATIONS SECURITY, 2017, : 1175 - 1191
  • [4] Catrina O, 2010, LECT NOTES COMPUT SC, V6052, P35, DOI 10.1007/978-3-642-14577-3_6
  • [5] Charles Z., 2020, arXiv
  • [6] On the Optimal Recovery Threshold of Coded Matrix Multiplication
    Dutta, Sanghamitra
    Fahim, Mohammad
    Haddadpour, Farzin
    Jeong, Haewon
    Cadambe, Viveck
    Grover, Pulkit
    [J]. IEEE TRANSACTIONS ON INFORMATION THEORY, 2020, 66 (01) : 278 - 301
  • [7] "Short-Dot": Computing Large Linear Transforms Distributedly Using Coded Short Dot Products
    Dutta, Sanghamitra
    Cadambe, Viveck
    Grover, Pulkit
    [J]. IEEE TRANSACTIONS ON INFORMATION THEORY, 2019, 65 (10) : 6171 - 6193
  • [8] HeteroSAg: Secure Aggregation With Heterogeneous Quantization in Federated Learning
    Elkordy, Ahmed Roushdy
    Avestimehr, A. Salman
    [J]. IEEE TRANSACTIONS ON COMMUNICATIONS, 2022, 70 (04) : 2372 - 2386
  • [9] Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures
    Fredrikson, Matt
    Jha, Somesh
    Ristenpart, Thomas
    [J]. CCS'15: PROCEEDINGS OF THE 22ND ACM SIGSAC CONFERENCE ON COMPUTER AND COMMUNICATIONS SECURITY, 2015, : 1322 - 1333
  • [10] Frigard A., 2021, PROC 17 INT S WIRELE, P1