Decentralized and robust privacy-preserving model using blockchain-enabled Federated Deep Learning in intelligent enterprises

2024-05-01

Applied Soft Computing

AI Summary

To address performance degradation caused by non-IID data and privacy leakage and trust bottlenecks arising from centralized architectures in federated deep learning (FDL), this paper proposes a blockchain-empowered decentralized FDL framework. The framework integrates zero-knowledge proofs, Practical Byzantine Fault Tolerance (PBFT) consensus, and differential privacy, deploying lightweight smart contracts on an Ethereum sidechain to ensure verifiable model updates, anonymous participant identities, and robust Byzantine node detection and removal. Evaluated on an industrial IoT dataset, the framework achieves 98.2% model accuracy, reduces communication overhead by 37%, attains >99.4% success rate against Byzantine attacks, and complies with GDPR requirements. This work is the first to deeply integrate verifiable computation with decentralized consensus in FDL systems, establishing a novel paradigm for secure, trustworthy, and efficient collaborative modeling in intelligent enterprise scenarios.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Security and Privacy: Blockchains and distributed ledgers
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing

Problem

• Enhances decentralized federated learning robustness.
• Addresses nonIID data distribution challenges.
• Secures against privacy and poisoning attacks.

Innovation

• Blockchain-enabled Federated Deep Learning
• Addresses nonIID data challenges
• Enhances security and privacy