Android Malware Detection Using API Calls and Permissions With Random Forest Classifier

Android malware poses a persistent and evolving threat to mobile security, considering its capability to compromise sensitive user data and evade traditional detection methods. While signature-based approaches remain effective against known threats, they often fail against obfuscated or zero-day malware. To address this challenge, this paper presents a machine learning-based framework for Android malware detection using a random forest classifier. The proposed method leverages static features (specifically API calls and permissions) to characterize application behaviour without requiring code execution. A comprehensive dataset was constructed by integrating labelled samples from CIC-AndMal2017, API details from Android Studio SDK (API levels 15–34), and legacy data from GitHub repositories (API levels 1–14), ensuring wide coverage of Android versions and threat patterns. Feature selection using correlation analysis and dimensionality reduction via principal component analysis was employed to optimize the model. The resulting classifier achieved a detection accuracy of 95%, demonstrating both high performance and computational efficiency. The strength of this work can be summarized in highlighting the value of multi-source data integration, informed feature selection, and ensemble learning in developing scalable, real-world solutions for Android malware detection. © 2013 IEEE.