Blockchain Empowered Federated Learning Ecosystem for Securing Consumer IoT Features Analysis.

Area of Science:

• Distributed Systems
• Machine Learning Security
• Internet of Things

Background:

• Consumer Internet of Things (CIoT) relies on centralized gateways and servers, posing security and privacy risks.
• Traditional Machine Learning (ML) centralized data analysis leads to data leakage and single points of failure.
• Federated Learning (FL) offers privacy but centralized aggregators still present control and data retrieval vulnerabilities.

Purpose of the Study:

• To propose a novel blockchain-controlled, edge intelligence federated learning framework for CIoT.
• To address the security and privacy concerns inherent in centralized ML and FL architectures.
• To enhance the distributed learning platform for secure and collaborative CIoT data analysis.

Main Methods:

• Developed a federated learning platform integrated with a blockchain network.
• Replaced the centralized aggregator with a decentralized blockchain for secure model aggregation.
• Ensured secure participation of gateway devices (GW) through blockchain's trustless, immutable, and anonymous properties.

Main Results:

• The proposed framework effectively enables collaborative learning in CIoT environments.
• Blockchain integration enhances security, privacy, and system resilience by eliminating single points of control.
• Experimental validation using the Stanford Cars dataset demonstrated the framework's effectiveness.

Conclusions:

• The blockchain-controlled federated learning framework provides a robust and secure solution for CIoT.
• Decentralization through blockchain mitigates privacy risks and enhances the reliability of distributed learning.
• This approach encourages wider user participation in CIoT data analysis while safeguarding sensitive information.