Blockchain-based Federated Learning with Limited Resources

Qian Wang, Haofei MengICCV 2022 · DOI

Abstract

Multiple clients involved in federated learning upload learning model parameters to a central server instead of raw data, which enhances the privacy of individual nodes and reduces the communication stress. However, the traditional architecture relies on a reliable and stable central server with crises of trust and single point failure. We merge the federated learning architecture with the blockchain to replace the centralized server and achieve a secure, fully decentralized blockchain federated learning. Focus on training scenarios with limited resources, we analyse system delays and design an algorithm adjusting local iterations to achieve better model performance. The performance of the proposed algorithm is evaluated in a simulated environment compared with training results for fixed local iterations. The results show that our method is well adapted in dynamic environments.
BibTeX
@inproceedings{wang2022blockchain,
  author = {Qian Wang and Haofei Meng},
  title = {{Blockchain-based Federated Learning with Limited Resources}},
  booktitle = {IEEE/CVF International Conference on Computer Vision},
  year = {2022},
  doi = {10.1109/cvidliccea56201.2022.9825317},
}

Loading related content...

Figure