Adaptive Data-driven Optimization using Transfer Learning for Resilient, Energy-efficient, Resource-aware, and Secure Network Slicing in 5G-Advanced and 6G Wireless Systems
Metadata[+] Show full item record
5G–Advanced is the next step in the evolution of the fifth–generation (5G) technology. It will introduce a new level of expanded capabilities beyond connections and enables a broader range of advanced applications and use cases. 5G–Advanced will support modern applications with greater mobility and high dependability. Artificial intelligence and Machine Learning will enhance network performance with spectral efficiency and energy savings enhancements. This research established a framework to optimally control and manage an appropriate selection of network slices for incoming requests from diverse applications and services in Beyond 5G networks. The developed DeepSlice model is used to optimize the network and individual slice load efficiency across isolated slices and manage slice lifecycle in case of failure. The DeepSlice framework can predict the unknown connections by utilizing the learning from a developed deep-learning neural network model. The research also addresses threats to the performance, availability, and robustness of B5G networks by proactively preventing and resolving threats. The study proposed a Secure5G framework for authentication, authorization, trust, and control for a network slicing architecture in 5G systems. The developed model prevents the 5G infrastructure from Distributed Denial of Service by analyzing incoming connections and learning from the developed model. The research demonstrates the preventive measure against volume attacks, flooding attacks, and masking (spoofing) attacks. This research builds the framework towards the zero trust objective (never trust, always verify, and verify continuously) that improves resilience. Another fundamental difficulty for wireless network systems is providing a desirable user experience in various network conditions, such as those with varying network loads and bandwidth fluctuations. Mobile Network Operators have long battled unforeseen network traffic events. This research proposed ADAPTIVE6G to tackle the network load estimation problem using knowledge-inspired Transfer Learning by utilizing radio network Key Performance Indicators from network slices to understand and learn network load estimation problems. These algorithms enable Mobile Network Operators to optimally coordinate their computational tasks in stochastic and time-varying network states. Energy efficiency is another significant KPI in tracking the sustainability of network slicing. Increasing traffic demands in 5G dramatically increase the energy consumption of mobile networks. This increase is unsustainable in terms of dollar cost and environmental impact. This research proposed an innovative ECO6G model to attain sustainability and energy efficiency. Research findings suggested that the developed model can reduce network energy costs without negatively impacting performance or end customer experience against the classical Machine Learning and Statistical driven models. The proposed model is validated against the industry-standardized energy efficiency definition, and operational expenditure savings are derived, showing significant cost savings to MNOs.
Table of Contents
Introduction -- A deep neural network framework towards a resilient, efficient, and secure network slicing in Beyond 5G Networks -- Adaptive resource management techniques for network slicing in Beyond 5G networks using transfer learning -- Energy and cost analysis for network slicing deployment in Beyond 5G networks -- Conclusion and future scope
Ph.D. (Doctor of Philosophy)