Adaptive Blockchain-Integrated Nonlinear Federated Learning Framework for Real-Time Intrusion Detection in IoT Fog Networks ABFL-RTID

Abstract

This research presents the Adaptive Blockchain-Integrated Nonlinear Federated Learning (ABFL-RTID) model, designed for real-time intrusion detection in IoT fog networks. The framework leverages nonlinear learning techniques, such as deep neural networks, to enhance detection capabilities in complex and dynamic network environments. Integrating blockchain ensures decentralized security, while federated learning preserves data privacy by enabling local model training on edge devices. The nonlinear models improve adaptability, accurately identifying sophisticated intrusion patterns while securely validating updates through blockchain. Simulations show the framework achieving 96.8% detection accuracy with response times of 150 ms, demonstrating superior scalability and adaptability under changing network conditions. This study provides a practical approach for building resilient and secure intrusion detection systems, enhancing data integrity and privacy without the delays of traditional, centralized models.