Date of Defense

18-4-2025 9:30 AM

Location

E1-1023

Document Type

Thesis Defense

Degree Name

Master of Science in Internet of Things

College

CIT

Department

Computer and Network Engineering

First Advisor

Prof. Najah

Keywords

Network Traffic Management, Distributed Reinforcement Learning (DRL), Large Language Models (LLMs), NF-TON-IOT Dataset, Machine Learning Classifiers, Random Forest, AdaBoost, C4.5, Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), RBF Kernel, Traffic Classification, Intrusion Detection

Abstract

The focus of this research is to explore collaborative network traffic management strategies using the Distributed Reinforcement Learning (DRL) and Large Language Models (LLMs) approaches. It emphasizes exploring a new tool for addressing network traffic by utilizing Distributed Reinforcement Learning (DRL) and Large Language Models (LLMs). This is achieved by utilizing self-organizing and self-directing techniques to optimize the network performance. Using the NF-TON-IOT dataset, various classifiers such as Random Forest, AdaBoost, C4. 5, Multi-Layer Perceptron (MLP), and SVM with an RBF kernel were tested for traffic classification and intrusion detection. Research recommends that DRL optimizes the complexity of the network by allowing agents to make decisions independently of the other agents; LLM optimizes the interaction between the agents in the network. By using the performance analysis, it has been explored that Random Forest and AdaBoost are more effective as compared to other classifiers that have been tested such as SVM with RBF kernel. However, the RBF kernel SVM has the drawbacks. The main drawback of the SVM RBF kernel is associated with its computational expenses and the longer training time especially in datasets of sizeable amount as NF-TON-IOT. This is not scalable and inefficient for real-time network traffic management scenario since fast adaptive responses are further needed.
The overall implementation of different confusion matrices for the accuracy check demonstrates that combining the DRL and LLM frameworks is conceivable for constructing a flexible and extensible network management architecture that can provide valuable guidance for the future development of network traffic management.

Share

COinS
 
Apr 18th, 9:30 AM

COLLABORATIVE NETWORK TRAFFIC MANAGEMENT STRATEGIES USING THE DISTRIBUTED REINFORCEMENT LEARNING AND LARGE LANGUAGE MODELS

E1-1023

The focus of this research is to explore collaborative network traffic management strategies using the Distributed Reinforcement Learning (DRL) and Large Language Models (LLMs) approaches. It emphasizes exploring a new tool for addressing network traffic by utilizing Distributed Reinforcement Learning (DRL) and Large Language Models (LLMs). This is achieved by utilizing self-organizing and self-directing techniques to optimize the network performance. Using the NF-TON-IOT dataset, various classifiers such as Random Forest, AdaBoost, C4. 5, Multi-Layer Perceptron (MLP), and SVM with an RBF kernel were tested for traffic classification and intrusion detection. Research recommends that DRL optimizes the complexity of the network by allowing agents to make decisions independently of the other agents; LLM optimizes the interaction between the agents in the network. By using the performance analysis, it has been explored that Random Forest and AdaBoost are more effective as compared to other classifiers that have been tested such as SVM with RBF kernel. However, the RBF kernel SVM has the drawbacks. The main drawback of the SVM RBF kernel is associated with its computational expenses and the longer training time especially in datasets of sizeable amount as NF-TON-IOT. This is not scalable and inefficient for real-time network traffic management scenario since fast adaptive responses are further needed.
The overall implementation of different confusion matrices for the accuracy check demonstrates that combining the DRL and LLM frameworks is conceivable for constructing a flexible and extensible network management architecture that can provide valuable guidance for the future development of network traffic management.