Title

SECURE AND ENERGY-EFFICIENT RESOURCE ALLOCATION IN NETWORK SLICING

Abstract

Abstract

The one-size-fits-all idea of the previous telecommunication generations is no longer
suitable for today’s applications. The current network system needs to satisfy the
Quality of Services of the different types of use cases such as Enhanced Mobile
Broadband, Ultra-Reliable, and Low Latency Communications and Massive Machine
Type Communications in the same physical infrastructure. 5G telecommunication
network is providing a solution to this problem: Network Slicing concept. Virtual
Network Functions (VNF) plays an essential role in the network slicing concept and
embedding these functions onto the network is an important task to do. As state-ofthe-art research focuses on allocating these functions on-network taking only energy
efficiency into consideration, this research brings a solution that considers the security aspects too. We propose a VNF placement strategy using an integer linear
programming (ILP) model for 5G network slicing under strict security requirements,
which optimizes energy consumption by the core network nodes. As an improvement to this approach, we also suggest using Deep Reinforcement Learning (DRL)
methods to provide a dynamic,energy-efficient, resilient, and secure resource allocation framework for network slicing. Specifically, 4 different Deep Reinforcement
Learning agents were trained to compare their results with the ILP-based framework
that we implemented. The result of the study shows that DRL-based methods provide faster allocation than ILP-based methods. Also, in this research, we compared
various state-of-the-art DRL methods to find a suitable algorithm for energy-efficient
resource allocation on stringent security constraints. Simulation results demonstrate
that the proposed model achieves significant power savings over a greedy approach
performing VNF placement under the same QoS and security constraints.

Supervisor(s)

Supervisor(s)

UMUT CAN GULMEZ

Date and Location

Date and Location

2022-09-02 11:00:00

Category

Category

MSc_Thesis