Federated Learning (FL) enables distributed machine learning without sharing raw data, but its reliance on a central aggregation server introduces critical vulnerabilities gradient inversion attacks, Byzantine poisoning, free-riding by rational participants, and single-point-of-failure risk. Blockchain has been proposed as a structural remedy, giving rise to the field of Blockchain-Enabled Federated Learning (BEFL). reviews exactly eight representative peer-reviewed BEFL systems published selected to cover four core challenge dimensions: privacy, security, scalability, and incentive design. compare each system across accuracy under data heterogeneity, formal privacy guarantees, Byzantine robustness, throughput, and communication efficiency. find that every reviewed system excels on one or two dimensions while underperforming on others, and that no single published system simultaneously resolves all four challenges. Based on this review identify four fundamental research gaps and conclude with a structured research agenda. This study also reveals that the base FL optimizer contributes more to model accuracy than any blockchain or privacy mechanism a finding with significant design implications.
Raman Dubey, A. Jain, Richa Sharma· International Conference on...· 0 citations
Training machine learning models across many institutions without moving raw data is attractive in theory, but the standard Federated Learning setup hands enormous power to one central aggregation server that participants must trust blindly. what happens when that trust assumption is violated and build a system, that does not require it. couples a four-layer distributed architecture with three stacked privacy mechanisms Local Differential Privacy to bound what any single gradient reveals, Secure Multi-Party Computation via the SecAgg protocol so the aggregation server never sees individual updates, and Tendermint-BFT blockchain consensus to create a tamper-evident audit log that no single party controls. Our main technical contribution is a formal proof that the combination of LDP and SecAgg satisfies end-to-end $(\varepsilon, \delta)$-differential privacy, including the SecAgg masking step cannot degrade the DP guarantee already earned by LDP noise injection. On CIFAR-10 with a ResNet-18 under three Dirichlet non-IID splits tested on real cluster hardware, reached 78.3% accuracy at $\varepsilon=0.83$ against six baselines including FedProx and SCAFFOLD-two standard non-IID benchmarks that prior blockchain FL work has consistently omitted.
Raman Dubey, A. Jain, Richa Sharma· International Conference on...· 0 citations