A federated learning-enhanced sharded blockchain framework for privacy-enhanced authentication in IoT E-learning systems

Abstract

The accelerated growth of IoT-facilitated e-learning ecosystems has introduced significant challenges for secure, scalable, and privacy-aware user authentication. Existing approaches face a fundamental trade-off: conventional blockchain systems often incur high latency and limited scalability, while centralized federated learning architectures may introduce privacy concerns and single points of failure. This study presents an authentication framework that integrates sharded blockchain architecture with federated learning to address these challenges. The proposed framework incorporates shard-based transaction processing, Byzantine fault tolerant (PBFT) consensus, resilient federated aggregation, and AES-GCM-encrypted model updates. Experimental results obtained under the adopted simulation settings indicate authentication accuracy of 94.03%, an AUC of 97.79%, an F1-score of 94.37%, and an EER of 5.97%. Compared with the selected blockchain-based baseline methods, the proposed framework demonstrated higher throughput and lower authentication latency under the evaluated workloads.