Detection of android and windows malware using machine learning

Malicious software (malware) continues to threaten global cybersecurity, causing extensive data breaches and financial losses. This study employs machine learning (ML) algorithms Decision Tree, Random Forest, and XGBoost to detect malware targeting Windows and Android platforms. Four diverse datasets were analyzed, and extensive preprocessing, feature selection, and hyperparameter optimization were performed. Results revealed that Random Forest achieved the highest accuracy of 99.37% on the TUNADROMD Android dataset, while XGBoost reached 85.94% on the large Android Malware Detection dataset after tuning. For Windows datasets, both Random Forest and XGBoost achieved 100% accuracy on smaller samples, with 87.0% and 86.98% respectively on the extensive Windows Portable Executable dataset. Feature selection improved Random Forest accuracy by 2–3%, while hyperparameter fine-tuning enhanced XGBoost by 8.94%. The study concludes that tree-based ensemble models, especially Random Forest and XGBoost, provide robust, generalizable, and interpretable solutions for malware detection across heterogeneous platforms.