Performance Evaluation of Machine Learning Techniques for Ransomware Detection: A Comparative Analysis

Abstract

In today's fast-paced digital world, application security has become an essential aspect for protecting organizations from increasing cyber-attacks and data breaches. Among these threats, ransomware is one of the most destructive malwares that attacks vulnerabilities of applications and systems through phishing, outdated software, and other insecure system configurations. As ransomware attacks are constantly evolving and traditional detection methods are not always capable of detecting attacks quickly and accurately, this study comparatively evaluate different machine learning (ML) algorithms for ransomware detection in terms of precision, recall, accuracy, F1-score, ROC-AUC, and logarithmic loss. The experimental results show that the Support Vector Machine (SVM) model shows better results than other classifiers. It has 97.73% precision, 96.00% recall, 98.00% accuracy, 97.00% F1-score, and the lowest logarithmic loss of 0.082, demonstrating superior effectiveness. The findings aim to strengthen application security, increase public trust in the digital ecosystem, and safeguard sensitive personal and institutional data.