Hybrid Blockchain-Based Resource Trading System for Federated Learning in Edge Computing

By training a machine learning algorithm across multiple decentralized edge nodes, federated learning (FL) ensures the privacy of the data generated by the massive Internet-of-Things (IoT) devices. To economically encourage the participation of heterogeneous edge nodes, a transparent and decentraliz...

Full description

Saved in:
Bibliographic Details
Published in:IEEE internet of things journal 2021-02, Vol.8 (4), p.2252-2264
Main Authors: Fan, Sizheng, Zhang, Hongbo, Zeng, Yuchen, Cai, Wei
Format: Article
Language:English
Subjects:
Algorithms
Auction
Blockchain
Computational modeling
Consortia
Cryptography
Edge computing
Federated learning
Hybrid systems
Internet of Things
Internet of Things (IoT)
Machine learning
Nodes
Peer-to-peer computing
Smart contracts
trade market
Training
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:By training a machine learning algorithm across multiple decentralized edge nodes, federated learning (FL) ensures the privacy of the data generated by the massive Internet-of-Things (IoT) devices. To economically encourage the participation of heterogeneous edge nodes, a transparent and decentralized trading platform is needed to establish a fair market among distinct edge companies. In this article, we propose a hybrid blockchain-based resource trading system that combines the advantages of both public and consortium blockchains. We design and implement a smart contract to facilitate an automatic, autonomous, and auditable rational reverse auction mechanism among edge nodes. Moreover, we leverage the payment channel technique to enable credible, fast, low-cost, and high-frequency payment transactions between requesters and edge nodes. Simulation results show that the proposed reverse auction mechanism can achieve the properties, including budget feasibility, truthfulness, and computational efficiency.
ISSN:2327-4662
2327-4662
DOI:10.1109/JIOT.2020.3028101