Enhancing Cybersecurity with Adaptive Firewalls: A Machine Learning Approach for Dynamic Rule Optimization

Lakshminarayanan, Adithyaprabakaran (2025) Enhancing Cybersecurity with Adaptive Firewalls: A Machine Learning Approach for Dynamic Rule Optimization. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (840kB) | Preview
Preview	PDF (Configuration Manual) Download (910kB) | Preview

Abstract

Conventional static, manually-maintained firewall rules have a hard time keeping up with the everchanging cyber-threat environment, resulting in high false-positive rates, lengthy mitigation times, and security vulnerabilities. The study develops a model of an adaptive firewall, which constantly learns based on real-time network traffic, and automatically optimizes its policies using supervised machine-learning. With this purpose in mind, two complementary intrusion-detection corpora, NSL-KDD of legacy patterns and UNSW-NB15 of modern attacks were joined, cleaned, label-encoded, statistically scaled and re-balanced with SMOTE-Tomek, and a 43-feature matrix with 261,676 training and 51,535 test flows was obtained. Eight candidate classifiers covering decision-tree logic, ensemble boosting, and deep neural architectures were designed, hyper-parameterised, and tested with stratified hold-out validation. The best-performing learner turned out to be gradient-boosted XGBoost, with 98.65 % accuracy, 0.989 macro F1-score, and sub-millisecond per-flow inference time and only 3.8 MB of memory-which meets real-time firewall latency requirements and can fit on edge devices with limited memory. Analysis of confusion-matrices established a false-positive rate of 1.8 % and a false-negative rate of 1.4 %, which alleviated operational alert fatigue without compromising detection sensitivity. The model and its scaler and encoders are serialised to be inserted into an adaptive rule-update micro-service, providing a reproducible blueprint of next-generation, self-optimising firewalls.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Pantridge, Michael UNSPECIFIED
Uncontrolled Keywords:	adaptive firewall; machine learning; XGBoost; NSL-KDD; UNSW-NB15; intrusion detection; dynamic rule optimisation; cyber security
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Cyber Security
Depositing User:	Ciara O'Brien
Date Deposited:	18 Aug 2026 16:43
Last Modified:	18 Aug 2026 16:43
URI:	https://norma.ncirl.ie/id/eprint/9539

Actions (login required)

View Item