Blockchain Federated Distillation with RL Incentives for Fair and Energy Efficient Edge IIoT
Year: 2026
Type: article
Abstract: The Industrial Internet of Things (IIoT) requires secure, privacy-preserving, and energy-efficient collaboration among heterogeneous edge devices.Traditional federated learning frameworks frequently suffer from high communication costs, decreased fairness, and limited scalability in resource-constrained environments.To address these issues, we proposed an adaptive incentive-driven blockchain-federated distillation framework that integrates blockchain coordination, knowledge distillation, and a Q-learning-based adaptive incentive mechanism.The blockchain layer, implemented using the Hyperledger Fabric and Raft consensus, ensures decentralized trust, auditability, and tamper-proof model exchanges.The incentive layer uses a fairness-aware Q-learning policy to dynamically assign rewards based on energy consumption, latency, accuracy, and Jain's Fairness Index, promoting equitable participation.Experiments on a heterogeneous IIoT testbed using Raspberry Pi and Jetson Nano devices showed that BFD achieved an 88.7% model accuracy (only 2.8% lower than centralized FL), 22% lower energy consumption, and a fairness index improvement of 29% compared with baseline methods.These results validate the proposed BFD as a scalable, fair, and energy-efficient solution for securing edge intelligence.
Authors: Ting Chai, Sung-Hwa Han, Seung-Soo Shin
Institution: Tongmyong University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7143680272
FWCI:
Cites:
Cited by:
Related to: