Blockchain Meets Federated Learning: A Comprehensive Survey on Smart Contract-driven Optimization for Aggregation, Security, and Scalability

Abstract

Blockchain enabled Federated Learning systems seek to unite the privacy guarantees of Federated Learning with the decentralized trust and auditability of blockchain networks. Although Federated Learning enables collaborative model training without sharing raw data, it remains exposed to malicious participants and lacks verifiable global coordination. The immutable ledger and consensus mechanisms of blockchain address these gaps while requiring programmable logic to orchestrate complex Federated Learning operations. In this comprehensive survey we demonstrate how smart contracts optimize blockchain enabled Federated Learning architectures by ensuring secure and transparent model aggregation, detecting and excluding malicious updates in real time, managing participant reputation dynamically, and distributing incentives to maintain honest participation. We present a taxonomy of smart contract designs and evaluate their impact on security, scalability, and gas cost tradeoffs. We compare practical implementations across diverse application domains. Finally, we highlight open challenges including gas cost optimization, privacy-preserving computation and cross silo interoperability and outline future directions for robust, scalable and autonomous blockchain enabled Federated Learning systems in trustless environments.