NUST Institutions Library Catalogue

The rapid expansion of the Internet of Things (IoT) has led to an unprecedented surge
in the number of connected devices, generating vast amounts of imbalance data and demanding
sophisticated management solutions. Traditional network architectures struggle
to cope with the dynamic and heterogeneous nature of IoT environments. Software
Defined Networking (SDN) emerges as a powerful paradigm to address these challenges
by decoupling the control plane from the data plane. Deep Learning (DL), a subset of
Machine Learning (ML), has demonstrated remarkable capabilities in processing and
analyzing large-scale data, positioning it as a powerful tool for improving SDN-IoT
systems. Researchers have made significant efforts to address SDN-IoT challenges by
applying ML and DL techniques. This thesis investigates the synergistic integration
of DL techniques within SDN-IoT frameworks. The primary objective is to develop
innovative DL-driven solutions that optimize network management, enhance security,
and improve overall system performance in IoT ecosystems. The first proposed method
employs a Generative Adversarial Network (GAN) model to accurately detect and classify
minor class attacks, achieving impressive accuracy and F1-score values of 94.44%
and 92.55%, respectively. The second technique utilizes a Deep Reinforcement Learning
(DRL) model in the SDN-IoT environment to achieve secure routing optimization. It is
evaluated using three benchmark metrics: throughput, latency, and the probability of
avoiding malicious nodes. This method demonstrated superior throughput and latency
performance compared to existing approaches in the literature.
This thesis makes several key contributions to the field of IoT networking. Firstly, it
provides a comprehensive analysis of the potential of DL to address critical challenges
in SDN-enabled IoT environments. Secondly, it introduces novel DL models tailored
for specific IoT applications, showcasing their effectiveness through extensive evaluations.
Thirdly, it highlights the importance of interdisciplinary approaches, combining advancements in networking and Artificial Intelligence (AI) to drive innovation in IoT
systems. Future research directions include exploring the integration of other emerging
technologies, such as Federated Learning and Blockchain, with DL and SDN-IoT
frameworks. Additionally, investigating the scalability of DL models in large-scale IoT
deployments and their adaptability to new and evolving network conditions remains a
promising area for further exploration.