MICHEAL, IFEANYICHUKWU MKPARU (2026) AN AI-POWERED NETWORK THREAT ASSESSMENT AND RECOMMENDATION SYSTEM. Other thesis, GODFREY OKOYE UNIVERSITY, ENUGU.
![[img]](https://eprints.gouni.edu.ng/style/images/fileicons/text.png) |
Text
Micheal Ifeanyichukwu Mkparu_main001.pdf Download (1MB) |
Abstract
The rise of digitalization in academic institutions has increased the vulnerability of campus networks to cybercriminal activities. Existing approaches to vulnerability scanning have been found to be dependent on the Common Vulnerability Scoring System (CVSS). The problem with CVSS is that it has been found to provide poor estimates regarding exploitability of vulnerabilities, leading to inefficiencies in security resource allocation. This has led to alert fatigue and late remediation of threats, especially in limited-resource settings such as Nigerian universities. The present paper concentrated on the design and development of an AI-Powered Network Threat Assessment and Recommendation System to prioritize, analyze, and interpret the vulnerability in the university network environment. The developed system uses an XGBoost binary classifier which has been trained on 155,852 real-world CVEs of a Kaggle CVE/EPSS/KEV enriched dataset using the CVSS score, criticality of assets, exploit likelihood (EPSS), and vulnerability age as predictors. The NLP has been used to come up with simplified, understandable vulnerability remediation suggestions. At the same time, the Explainable AI (XAI) approach using SHAP (Shapley Additive Explanation) makes the model's predictions transparent. The system harvests the vulnerability information from Nmap network scan, which is enriched with the help of live threat intelligence collected from the National Vulnerability Database (NVD), FIRST.org EPSS API, and CISA Known Exploited Vulnerabilities (KEV) database. Finally, a web-based dashboard has been developed in the Flask, Bootstrap, and Chart.js technologies. After the training process, the obtained AUC-ROC of the trained classifier on the holdout set was 0.946 (5-fold cross-validation AUC of 0.952 ± 0.009) and the recall was 80.6%. The system was able to get 90% Precision@50, meaning that 90% of the 50 highest-priority AI-ranks were confirmed exploited vulnerabilities, whereas for the CVSS-only ranking it was just 22%.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | H Social Sciences > H Social Sciences (General) |
| Divisions: | Faculty of Computing And Information Technology (FACIT) > School of Chemistry |
| Depositing User: | CHIBUEZE EZE |
| Date Deposited: | 24 Jul 2026 08:30 |
| Last Modified: | 24 Jul 2026 08:30 |
| URI: | http://eprints.gouni.edu.ng/id/eprint/5897 |
Actions (login required)
 |
View Item |
