Integrated fuzzy trust management with federated machine learning approaches for enhancing intrusion detection in IoT networks

As the Internet of Things (IoT) expands, our environment is increasingly composed of vast networks of interconnected devices that generate large volumes of data. These smart devices improve efficiency and enhance the quality of human life; however, their diversity and scale create significant challenges in terms of trustworthiness and security. Traditional security solutions often struggle to handle the dynamic and complex threat landscape of IoT systems. To address these challenges, we propose a novel framework that integrates fuzzy trust management (FTM), the synthetic minority oversampling technique (SMOTE), and a federated support vector machine model (FTS-FedSVM). Fuzzy trust management enables accurate evaluation of trust by considering multiple factors, such as device behavior, network activity, and reputation, providing the adaptability needed for evolving threats. FedSVM allows distributed devices to collaboratively train models without sharing raw data, which reduces computational overhead while protecting privacy. This cooperative strategy ensures that even resource-constrained IoT devices benefit from the collective knowledge of the network. Extensive experimental evaluations demonstrate the effectiveness of various federated learning models. In contrast, the proposed Fed-SVM framework significantly outperforms these models, achieving an impressive 99.03% accuracy in binary classification (benign vs. malicious nodes) and 91.30% accuracy in multi-class attack detection, highlighting its robustness and superior learning capability compared to Federated Random Forest and Federated Decision Tree. It effectively mitigates threats such as botnets, ransomware, and network sniffing in heterogeneous IoT environments. Moreover, the framework reduces computational power consumption by approximately 80%, demonstrating its suitability for sustainable IoT security.