Detecting Vulnerabilities in Senayan Library Management System Using Machine Learning

Abstract

Vulnerabilities in the open-source Senayan Library Management System (SLiMS) can impact thousands of libraries globally. However, manual code reviews to uncover undisclosed weaknesses are time-consuming. This study proposes an automated machine learning based static analysis approach to detect vulnerabilities in SLiMS PHP code. A dataset of 22,877 samples with cross-site scripting, SQL injection and other weakness types is used to train classification models. Feature engineering extracted discriminative attributes indicating flaws while vectorization structured the code samples. A Multinomial Naïve Bayes model achieved 97% accuracy in detecting security issues. Three critical vulnerabilities were identified in SLiMS v9.6.1 which were responsibly disclosed and fixed. The findings demonstrate feasibility of applying machine learning to boost open-source security. Further enhancements through custom datasets and deep neural networks can aid real-time vulnerability discovery.