A Secure and Energy-Efficient Cross-Layer Framework for Internet of Things Networks

Abstract

In resource-constrained environments, achieving the optimal balance between security and energy efficiency remains a fundamental challenge in the design of the Internet of Things (IoT) networks. This thesis proposes and reports on a novel cross-layer framework spanning the Application, Network, and Sensor Layers that is both secure and energy-efficient. The methodology integrates comprehensive simulations, real-world testbeds and machine learning (ML) models to design and validate the proposed architecture. A runtime adaptive cryptographic system employing lightweight encryption algorithms Speck, and Present with dynamic round reduction is developed to minimize energy consumption without compromising cryptographic strength. To further enhance threat resilience, ML based intrusion detection is incorporated across all layers, utilizing models such as decision trees and Long Short-Term Memory (LSTM) networks, resulting in significantly improved anomaly detection accuracy. Empirical results from Contiki/Cooja and NS-3 simulations, alongside hardware evaluations, confirm that the framework improves packet delivery, reduces latency, and enhances power efficiency. Overall, the cross-layer architecture demonstrates robust scalability, resilience to cyberattacks, and practical suitability for sustainable IoT deployments in real-world, resource-constrained scenarios. This research addresses gaps in prior single-layer security models by proposing a novel, integrated design and sets the foundation for future works.