Machine Learning Driven Security Vulnerability Detection in Large-Scale Enterprise Software Systems

View at publisher Summarise with VIRAT
Authors: Tarun Kalwani
Cited by: 0
References: 15

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

With increasing scale and complexity, traditional manual and signature-based vulnerability detection methods in software systems become ineffective. Modern applications expose large attack surfaces, and thus early identification of security weaknesses needs to be handled intelligently and automatically. This paper presents a machine learning-driven framework for classifying source code snippets as vulnerable or non-vulnerable before deployment. Using a curated dataset of 439 manually annotated C++ code samples, the study investigates three supervised learning models, namely, Logistic Regression, Random Forest, and a Deep Neural Network, after transforming code into TF-IDF feature vectors through tokenization and pre-processing. Experimental results show that among others, the Random Forest classifier achieves the best overall performance with 88.6% accuracy, 90.7% precision, and an 88.6% F1-score, outperforming both traditional baseline scanners and other ML models. These findings emphasize the role of ensemble methods for dealing with high-dimensional code representations and indicate that machine learning can significantly enhance the software development lifecycle by reducing false positives, improving consistency in detection, and enabling earlier remediation of vulnerabilities. The work underlines the practical feasibility of ML-based detection of software vulnerabilities and sets the basis for future works based on state-of-the-art code-aware representations, such as AST-based models, GNNs, and transformer-based architectures.

Fields of Study & Subjects

Software Engineering Research Web Application Security Vulnerabilities Software Testing and Debugging Techniques

Publication details

Journal / source
International Conference on Image Processing and Capsule Networks (ICIPCN)
Pages
1405–1410
Publication date
27 January 2026
Language
EN