All theses

Permanent URI for this collection

http://62.72.18.19:4000/handle/123456789/6

Browse

Browsing All theses by Author "Adomako, N.J."

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    A Comparative Study of Classical and Quantum Support Vector Machines for Distributed Denial of Service Attack Detection on Network Traffic Data
    (UENR, 2025-12) Adomako, N.J.
    The escalating frequency and sophistication of Distributed Denial of Service (DDoS) attacks pose a significant threat to the integrity and availability of modern digital infrastructures, particularly in developing countries like Ghana. Traditional Intrusion Detection Systems (IDS) which are reliant on static rule sets, struggle to detect zero-day and high-volume attacks, necessitating more adaptive and intelligent solutions. Classical machine learning models such as Support Vector Machines (SVMs) have demonstrated strong classification capabilities in network traffic analysis; however, they face limitations in scalability, latency, and efficiency when applied to high￾dimensional or encrypted data streams. This study presents a comparative analysis of classical SVM and Quantum Support Vector Machine (QSVM) models for Distributed Denial of Service attack detection using the publicly available Software Defined Network Distributed Denial of Service dataset. The classical Support Vector Machines wasimplemented using Scikit-learn’s Support Vector Classifier class with a Radial Basis Form kernel, while the Quantum Support Vector Machine (QSVM) leveraged Quantum Information Science Kit’s Quantum Support Vector Classifier (QSVC) class with a ZZFeatureMap and quantum kernel evaluated on simulated backends. Both models underwent identical preprocessing, including data cleaning, one-hot encoding, feature scaling, and dimensionality reduction via Principal Component Analysis (PCA). Experimental results revealed that the classical Support Vector Machine outperformed Quantum Support Vector Machine in terms of accuracy, precision, recall, and Receiver Operating Characteristic curve performance, achieving 95.7% accuracy compared to Quantum Support Vector Machine’s 82.5%. However, Quantum Support Vector Machine demonstrated potential scalability benefits and theoretical advantages in kernel expressivity. The study highlights the practical limitations of current quantum simulators and emphasizes the need for hardware advancements to fully realize Quantum Support Vector Machine’s promise in cybersecurity contexts. By bridging classical and quantum approaches, this research contributes empirical insights to the emerging field of Quantum Cybersecurity and offers guidance for deploying hybrid detection systems in real-world network environments. The findings are particularly relevant for cybersecurity practitioners, researchers, and policymakers seeking next-generation solutions to evolving cyber threats.