AuditFL uses blockchain to make federated learning auditable

Researchers have developed AuditFL, a novel system designed to enhance the transparency and accountability of open federated learning. AuditFL utilizes blockchain technology, specifically Solidity contracts on an Ethereum-compatible Anvil chain, to manage client admissions, track model updates, and handle rewards without removing the central aggregator. This approach ensures that the aggregator's decisions are attributable, replayable, and economically constrained, while keeping the intensive training and aggregation processes off-chain. Experiments on the MNIST dataset with non-IID splits demonstrated that AuditFL maintains the learning trajectory of standard federated learning, even under significant failure conditions like client dropout or invalid updates, with minimal impact on final accuracy. AI

IMPACT Enhances transparency and audibility in federated learning systems, potentially increasing trust and adoption.

RANK_REASON Academic paper detailing a new system for federated learning.

Read on Hugging Face Daily Papers →

• anvil
• arXiv
• AuditFL
• Ethereum
• Ethereum Virtual Machine
• federated learning
• FLTA 2026
• Hugging Face
• MNIST database
• Solidity

• paper
• infra

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →