Vol. 2, Issue 1, Part A (2025)

The exponential growth of the Internet of Things (IoT) has revolutionized connectivity and automation but has also amplified concerns over data security, integrity, and privacy. Conventional centralized machine learning (ML) models in IoT systems are increasingly inadequate for handling distributed data and ensuring trust among edge devices. This study introduces a Blockchain-Assisted Federated Learning (BA-FL) framework that integrates decentralized ledger technology with collaborative learning to enhance security and model reliability in IoT networks. The system employs a permissioned blockchain to record cryptographically hashed model updates, enforce consensus validation using the Practical Byzantine Fault Tolerance (PBFT) algorithm, and ensure tamper-proof provenance of learning transactions. Two benchmark datasets, NSL-KDD and BoT-IoT, were used to evaluate the performance of the proposed model against conventional federated learning (FL). Experimental results revealed significant improvements in detection performance—achieving an average F1-score of 0.955 on NSL-KDD and 0.968 on BoT-IoT—and a drastic reduction in model-poisoning attack success rates from approximately 20% to under 7%. Statistical analysis confirmed that these gains were significant (p < 0.001). Although the blockchain integration introduced moderate overhead in latency and energy consumption, the trade-off remained within acceptable operational limits for real-time IoT applications. The findings demonstrate that blockchain-assisted ML frameworks can effectively mitigate security vulnerabilities, provide immutable audit trails, and maintain scalability across heterogeneous IoT networks. The study concludes that the BA-FL model represents a practical and scalable solution for deploying secure, transparent, and trustworthy IoT intelligence, establishing a foundation for resilient edge analytics and autonomous digital infrastructures of the future.