A High-Performance Malware Detection Using Extra Tree Classifier and Logistic Regression on TUNADROMD Dataset

Precise malware identification is essential for enhancing cybersecurity frameworks against advancing threats. This study introduces an enhanced malware detection framework utilizing machine learning methods on the TUNADROMD data set from Kaggle, which comprises 4,465 instances and 242 instances. A selection of 23 high-impact attributes was identified through feature relevance analysis to enhance classification accuracy reduce computing complexity. Two models — Extra Tree Classifier and Logistic Regression — were assessed for their classification efficacy. The Extra Tree Classifier exhibited enhanced accuracy, attaining 97.42% in training and 97.23% in testing, surpassing the Logistic Regression model, which recorded 94.84% and 93.67%, respectively. The Extra Tree Classifier has superior predictive performance and reduced computational expense compared to current methodologies, positioning it as a dependable instrument for real-time malware detection. This work emphasizes the capability of machine learning to offer scalable and resilient solutions for malware analysis and detection, establishing a basis for future progress in automated threat mitigation.