Zero-Knowledge Proof-Based Consensus for Blockchain-Secured Federated Learning

TL;DR

This paper addresses privacy and trust challenges in blockchain-secured federated learning by introducing ZKPoT, a zero-knowledge proof of training-based consensus that verifies client contributions without exposing private data or models. It combines zk-SNARKs (specifically Groth16), content-addressed storage via IPFS, and a trusted task publisher to enable non-interactive, verifiable model evaluation while keeping communication and on-chain storage efficient. The authors design a complete FL-on-blockchain system with novel transaction/block formats, quantization techniques for zk-arithmetic, and a federation-wide leader election based on validated model performance, achieving resilience to Byzantine faults and preserving model accuracy without the trade-offs of differential privacy. Empirical results on CIFAR-10 and MNIST show robust privacy protection, scalable verification, and comparable or superior accuracy relative to DP-based baselines, highlighting practical potential for decentralized, privacy-preserving FL in large blockchain networks.

Abstract

Federated learning (FL) enables multiple participants to collaboratively train machine learning models while ensuring their data remains private and secure. Blockchain technology further enhances FL by providing stronger security, a transparent audit trail, and protection against data tampering and model manipulation. Most blockchain-secured FL systems rely on conventional consensus mechanisms: Proof-of-Work (PoW) is computationally expensive, while Proof-of-Stake (PoS) improves energy efficiency but risks centralization as it inherently favors participants with larger stakes. Recently, learning-based consensus has emerged as an alternative by replacing cryptographic tasks with model training to save energy. However, this approach introduces potential privacy vulnerabilities, as the training process may inadvertently expose sensitive information through gradient sharing and model updates. To address these challenges, we propose a novel Zero-Knowledge Proof of Training (ZKPoT) consensus mechanism. This method leverages the zero-knowledge succinct non-interactive argument of knowledge proof (zk-SNARK) protocol to validate participants' contributions based on their model performance, effectively eliminating the inefficiencies of traditional consensus methods and mitigating the privacy risks posed by learning-based consensus. We analyze our system's security, demonstrating its capacity to prevent the disclosure of sensitive information about local models or training data to untrusted parties during the entire FL process. Extensive experiments demonstrate that our system is robust against privacy and Byzantine attacks while maintaining accuracy and utility without trade-offs, scalable across various blockchain settings, and efficient in both computation and communication.

Figures (11)

• Figure 1: The architecture of ZKPoT-based Blockchain-secured Federated Learning system
• Figure 2: Zero-knowledge proof generation process
• Figure 3: Global model accuracy comparing our method to DP methods with different noise levels on the CIFAR10 dataset
• Figure 4: Global model accuracy comparing our method to DP methods with different noise levels on the MNIST dataset
• Figure 5: Effectiveness of membership inference attack during the model verification phase on the CIFAR10 dataset. ZKPoT conceals model parameters from unverified clients, reducing the attack to near-random guessing