Abstract:Supported by cloud computing, Federated Learning (FL) has experienced rapid advancement, as a promising technique to motivate clients to collaboratively train models without sharing local data. To improve the security and fairness of FL implementation, numerous Blockchain-empowered Federated Learning (BFL) frameworks have emerged accordingly. Among them, consensus algorithms play a pivotal role in determining the scalability, security, and consistency of BFL systems. Existing consensus solutions to block producer selection and reward allocation either focus on well-resourced scenarios or accommodate BFL based on clients' contributions to model training. However, these approaches limit consensus efficiency and undermine reward fairness, due to involving intricate consensus processes, disregarding clients' contributions during blockchain consensus, and failing to address lazy client problems (malicious clients plagiarizing local model updates from others to reap rewards). Given the aforementioned challenges, we make the first attempt to design a joint solution for efficient consensus and fair reward allocation in heterogeneous BFL systems with lazy clients. Specifically, we introduce a generalizable BFL workflow that can address lazy client problems well. Based on it, the global contribution of BFL clients is decoupled into five dominant metrics, and the block producer selection problem is formulated as a reward-constraint contribution maximization problem. By addressing this problem, the optimal block producer that maximizes global contribution can be identified to orchestrate consensus processes, and rewards are distributed to clients in proportion to their respective global contributions. To achieve it, we develop a Context-aware Proof-of-Contribution consensus algorithm named CPoC to reach consensus and incentive simultaneously, followed by theoretical analysis of lazy client problems and privacy issues. Empirical results on widely-used datasets demonstrate the effectiveness of our design in improving consensus efficiency and maximizing global contribution.

Proof-of-Collaborative-Learning: A Multi-winner Federated Learning Consensus Algorithm

A Platform-Free Proof of Federated Learning Consensus Mechanism for Sustainable Blockchains

Proof of Federated Learning: A Novel Energy-recycling Consensus Algorithm

FedChain: An Efficient and Secure Consensus Protocol based on Proof of Useful Federated Learning for Blockchain

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

Context-aware Consensus Algorithm for Blockchain-empowered Federated Learning

Proof of Learning (PoLe): Empowering Machine Learning with Consensus Building on Blockchains

A verifiable and privacy-preserving blockchain-based federated learning approach

Proof-of-Contribution-Based Design for Collaborative Machine Learning on Blockchain

Blockchain-Based Federated Learning: Incentivizing Data Sharing and Penalizing Dishonest Behavior

Distributed and Secure Federated Learning for Wireless Computing Power Networks

Proof of Learning: Towards a Practical Blockchain Consensus Mechanism Using Directed Guiding Gradients (student Abstract)

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

Secure and Efficient Decentralized Federated Learning with Data Representation Protection

Towards blockchain-enabled decentralized and secure federated learning

Resilient Consensus Sustained Collaboratively

Proof of Fairness: Dynamic and Secure Consensus Protocol for Blockchain

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

Secure Collaborative Learning in Mining Pool Via Robust and Efficient Verification

Enhancing the performance of permissionless blockchain networks through randomized message-based consensus algorithm

PoV: An Efficient Voting-Based Consensus Algorithm for Consortium Blockchains