Key points are not available for this paper at this time.
Operational Federated Learning (FL) across multiple operators in Open Radio Access Network (O-RAN) settings demands trustworthy coordination. We introduce FL-DApp, a blockchain-based, open-source Decentralized Application (DApp) for lightweight, reputation-driven client selection. Aligned with the O-RAN architecture, FL-DApp exposes a minimal, portable on-chain interface on Ethereum Virtual Machine (EVM) networks, supported by off-chain adapters that (i) register participants, (ii) collect and verify per-round validation metrics, and (iii) compute and store reputations. Model training and aggregation remain off-chain to minimize computational overhead. To support reproducibility and adoption, FL-DApp includes a scriptable workflow and generates structured outputs and CSV logs. The reference implementation targets the Polygon Amoy testnet and is portable across EVM-compatible layers. We evaluate registration, metric submission, reputation update, and finalization under realistic conditions. Around 86% of batched reputation commits are confirmed within seven seconds, enabling seconds-scale cycles aligned with the non-real-time RAN Intelligent Controller (Non-RT RIC) in the Service Management and Orchestration (SMO) framework. FL-DApp thus offers a practical and reproducible foundation for transparent, decentralized FL in multi-operator O-RAN environments.
Javed et al. (2026) studied this question.