Content area
The rapid expansion of the Internet of Things (IoT) has introduced significant security vulnerabilities due to the resource-constrained nature of IoT devices and their exposure to cyber threats. Traditional security solutions are often infeasible due to the high computational and storage demands they impose. This dissertation presents a lightweight, AI-driven security framework that enhances IoT network resilience by integrating feature selection, ensemble learning, and federated transfer learning while maintaining data privacy and minimizing computational overhead.
The proposed framework consists of three primary components: Feature Selection for Intrusion Detection, which optimizes performance by reducing redundant data and improving detection accuracy with minimal resource consumption; Ensemble Learning with Adaptive Model Selection, designed to enhance threat detection while conserving energy through efficient machine learning models; Federated Transfer Learning for IoT Security which enables collaborative model training across distributed devices without requiring raw data transfer, ensuring privacy preservation and real-time adaptability.
Experimental evaluations using benchmark IoT security datasets demonstrate that the proposed framework achieves up to 99.97% accuracy while significantly reducing computational costs compared to conventional security mechanisms. Furthermore, the federated learning approach mitigates privacy risks by preventing direct data exchanges among IoT nodes. The findings highlight the feasibility of scalable, privacy-preserving, and resource-efficient intrusion detection for IoT networks.
This research contributes to the advancement of AI-driven cybersecurity solutions, providing a robust and adaptable approach to safeguarding IoT environments from evolving threats. By addressing key challenges in IoT security, this work paves the way for future developments in smart, efficient, and self-adaptive security mechanisms for large-scale deployments.
