Abstract:Blockchain based federated learning is a distributed learning scheme that allows model training without participants sharing their local data sets, where the blockchain components eliminate the need for a trusted central server compared to traditional Federated Learning algorithms. In this paper we propose a softmax aggregation blockchain based federated learning framework. First, we propose a new blockchain based federated learning architecture that utilizes the well-tested proof-of-stake consensus mechanism on an existing blockchain network to select validators and miners to aggregate the participants' updates and compute the blocks. Second, to ensure the robustness of the aggregation process, we design a novel softmax aggregation method based on approximated population loss values that relies on our specific blockchain architecture. Additionally, we show our softmax aggregation technique converges to the global minimum in the convex setting with non-restricting assumptions. Our comprehensive experiments show that our framework outperforms existing robust aggregation algorithms in various settings by large margins.

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are several key challenges in the existing Federated Learning (FL) frameworks: 1. **Single Point of Failure**: Traditional federated learning relies on a central server to coordinate the training process, which leads to the problem of single - point failure. If the central server is affected by external attacks or network failures, the entire federated learning process may fail. 2. **Limitations of the Reward Mechanism**: Many Blockchain - based Federated Learning (BCFL) schemes have created their own reward mechanisms, and these rewards are of little value outside the system, thus affecting the enthusiasm of participants. 3. **Limited Defense Against Malicious Attacks**: The existing BCFL frameworks have limited defense capabilities against malicious attacks. Especially in the case of uneven participant dataset sizes and distributions, the existing robust aggregation methods perform poorly in specific settings and are difficult to generalize to more general situations. To solve the above problems, the paper proposes a new blockchain - based federated learning framework - SABFL (Softmax Aggregation on Blockchain - based Federated Learning). The main contributions of SABFL include: 1. **Utilizing Existing Blockchain Networks**: SABFL relies on existing blockchain networks (such as Ethereum) to handle the selection of validators and miners and reward management. This method not only encourages more participants to join, but also increases the attractiveness of rewards by using widely - accepted cryptocurrencies (such as Ether). 2. **Softmax - based Robust Aggregation Technique**: SABFL introduces a softmax - based aggregation technique, which can effectively deal with the heterogeneity of dataset sizes and distributions and performs well in the presence of malicious attacks. Specifically, SABFL calculates the approximate overall loss value of each participant and uses the softmax function to convert these loss values into weights, thereby achieving a weighted average of model updates. 3. **Convergence Guarantee**: The paper proves that in the convex optimization setting, the softmax aggregation technique of SABFL can converge to the global minimum, which provides theoretical support for the effectiveness of SABFL. 4. **Experimental Verification**: Through extensive experiments, the paper shows the superior performance of SABFL in various settings. Especially in the case of uneven participant dataset sizes and distributions and the presence of malicious attacks, the performance of SABFL is significantly better than that of the existing robust aggregation methods. In conclusion, SABFL solves the problems of single - point failure, limitations of the reward mechanism, and insufficient defense against malicious attacks in traditional federated learning by combining blockchain technology and the softmax aggregation method, providing a more secure, efficient, and robust solution for federated learning.

Robust softmax aggregation on blockchain based federated learning with convergence guarantee

BlockDFL: A Blockchain-based Fully Decentralized Peer-to-Peer Federated Learning Framework

A Blockchain-based federated learning framework for secure aggregation and fair incentives

Towards blockchain-enabled decentralized and secure federated learning

Secure and Efficient Decentralized Federated Learning with Data Representation Protection

Wait or Not to Wait: Evaluating Trade-Offs between Speed and Precision in Blockchain-based Federated Aggregation

A Blockchain Based Decentralized Gradient Aggregation Design for Federated Learning

A Secure and Fair Federated Learning Framework Based on Consensus Incentive Mechanism

DFL: High-Performance Blockchain-Based Federated Learning

Privacy-Preservation Robust Federated Learning with Blockchain-based Hierarchical Framework.

BRFL: A blockchain-based byzantine-robust federated learning model

Fair and Robust Federated Learning via Decentralized and Adaptive Aggregation based on Blockchain

Robust Aggregation for Federated Learning

Time-Efficient Blockchain-Based Federated Learning

Straggler Mitigation and Latency Optimization in Blockchain-based Hierarchical Federated Learning

When Blockchain Meets Asynchronous Federated Learning

Blockchain-Enabled Asynchronous Federated Learning in Edge Computing

When Federated Learning Meets Blockchain: A New Distributed Learning Paradigm

Transparent Contribution Evaluation for Secure Federated Learning on Blockchain

Asynchronous Federated Learning System Based on Permissioned Blockchains

Robust Model Aggregation for Heterogeneous Federated Learning: Analysis and Optimizations