Date of Award

Winter 3-1-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Cyberspace Engineering

First Advisor

William Bradley Glisson

Abstract

As the number of online users grows exponentially, the number and severity of cyber threats escalate, urgently requiring advancements in real-time network modeling and response. Swiftly predicting and analyzing network traffic is crucial for effective network monitoring and control, preventing cyber breaches, and maintaining healthy network functionality. This research presents a novel approach to real-time modeling based on analyzing evolving properties and patterns in a dynamical network system using a hybrid analog-digital computer. An analog computer was utilized as a co-processor to compute differential equations that model the Transmission Control Protocol (TCP) window size. A comparative analysis was conducted between the digital model, using Euler’s numerical integration method, and the hybrid model. Based on the Median Absolute Percent Error (MedAPE) statistical metric, the digital model attains a 98.87% accuracy in predicting TCP window size, while the hybrid model achieves 91.85% accuracy, excluding the precision loss from the pre-processed input signals. The hybrid model demonstrated a 570% improvement in execution time performance compared to the digital model, proving superior speed with minimal sacrifice on accuracy. The analysis further compares linear and nonlinear dynamical approaches, including Model Predictive Control (MPC), frequency-domain techniques, and Ordinary Differential Equations (ODEs), underscoring the hybrid system’s speed advantage, particularly for complex, real-time applications. The findings establish hybrid analog-digital systems as a potent alternative for dynamic network traffic modeling, providing a foundation for broader applications in nonlinear dynamical systems.

Download

Share

COinS