Comprehensive Ransomware Detection: Optimization of Feature Selection through Machine Learning Algorithms and Explainable AI on Memory Analysis

Abstract

The increase in ransomware attacks has underscored the need for ro bust cybersecurity measures. To combat these sophisticated threats, organiza tions must implement strong defenses, including cutting-edge technologies like machine learning to detect early signs of ransomware in their systems. This paper presents a comprehensive study on ransomware detection, highlighting the integration of machine learning algorithms and explainable artificial intelli gence (XAI) techniques to enhance the transparency and reliability of predictive models in this field. Our focus relies on optimizing features within the CIC-MalMem-2022 dataset, which contains various memory-based malware sam ples. We also use the decision tree algorithm to identify influential features and uses the SHAP model for transparent decision-making. The results demonstrate that the algorithms can efficiently detect ransomware using only five optimized features.