Blockchain-Based Decentralized Federated Learning Model

Abstract

As a decentralized technology, blockchain has the characteristics of traceability and immutability. Using blockchain technology, each model update of federation information can be associated with each participant, and the traceability of transactions in the blockchain can be used to detect fraudulent transactions that attempt to tamper with data. At the same time, the blockchain is used to realize the decentralization of the system, which strengthens the fault tolerance and attack resistance of the system. Based on this, this chapter designs and implements a blockchain architecture to solve the problem of safe and trusted data sharing among participants in the decentralized federated learning model. A blockchain architecture is designed and implemented for decentralized federated learning, using the blockchain to store the model parameter information of participants in each round of training; in view of the shortcomings of existing consensus algorithms, combined with decentralized federated learning detection model, A new consensus algorithm - “High Performance Byzantine Fault Tolerance Based on Model Accuracy” consensus algorithm is proposed.