Novel resource provisioning and lightweight security protocols for IoT edge networks

This Ph.D. thesis introduces a novel dynamic resource allocation framework tailored for Edge Computing (EC) in Internet of Things (IoT) networks, addressing the pressing challenges posed by resource limitations and escalating user demands. Edge-driven IoT networks, characterized by their reliance on locally available computational resources from a heterogeneous ensemble of devices such as sensors, vehicles, and mobile phones, present unique challenges. These resources, in contrast to their cloud counterparts, exhibit inherent variability in terms of processing power, distribution, and operating system diversity. Moreover, their connectivity is subject to fluctuations, including failures, intermittent connections, and unpredictable network entry and exit events, rendering the EC network inherently dynamic. The inadequacy of existing solutions to effectively manage the dynamic nature of resource availability at the edge underscores the necessity for a resource allocation framework capable of adapting to these dynamic conditions. To this end, we propose a dynamic resource allocation framework that dynamically assigns computational and network resources. This framework aims to minimize average service delays and achieve resource utilization balance at the edge. To realize this objective, two resource allocation models are developed using TensorFlow: a classification-based approach and a regression-based approach. Experimental results in dynamic environments demonstrate remarkable performance improvements, with the regression model achieving an 87% task completion rate within specified time constraints and the classification model achieving 56%. To underscore the practicality and efficiency of our proposed framework, two real-world use cases are explored. The first use case deals with the detection of spoofing attacks in autonomous vehicles (AVs) using Shadow Analyzer, a technique that identifies ghost object attacks with reduced 2D data derived from 3D point cloud information. The second use case focuses on the implementation of homomorphic encryption for secure communication, presenting a novel distributed approach to Fully Homomorphic Encryption (FHE)-based data processing. To validate the applicability and efficiency of our framework, extensive simulation experiments are conducted across various scenarios and operational conditions on a hardware testbed. These experiments yield promising results, establishing the viability of our dynamic resource allocation framework in addressing the dynamic challenges posed by resource availability at the edge in IoT networks.