Development of an AI-Driven Vulnerability Detection and Mitigation System
Year: 2026
Type: article
Abstract: This paper presents the development and evaluation of an AI-driven vulnerability detection and mitigation system for web applications. The system combines targeted vulnerability scanning with a machine-learning component to improve user control, assessment efficiency, and reporting. A Design Science Research approach combined with experimental evaluation was adopted. The prototype integrates a web crawler, interactive endpoint selection, vulnerability-specific scanners, logistic regression classification, and a reporting mechanism that provides severity information and mitigation recommendations. Testing was conducted in controlled and ethically authorized environments. The reported evaluation achieved detection rates of 90% for SQL Injection, 80% for Cross-Site Scripting (XSS), 100% for Cross-Site Request Forgery (CSRF), and 80% for Directory Traversal. Targeted scanning also reduced scan time from 12 to 5 minutes on a 20-page test site and from 45 to 20 minutes on a site containing more than 100 endpoints. The findings show that a hybrid workflow combining conventional vulnerability scanners, interactive endpoint selection, and machine learning can provide a practical approach to focused web-application security assessment. The prototype is nevertheless limited in detecting some complex vulnerabilities and requires broader datasets and further validation.
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204486064