Machine Learning-based Malware Detection on Android Operating System using AdaBoost Algorithm and ReliefF Feature Selection Method

Technological advancements occur rapidly from year to year. Currently, we experience the benefits of this rapid technological development, including the widespread use of smartphones in our daily lives. However, without robust security measures, smartphones can lead to losses for users and associated individuals. Additionally, as malware evolves, detecting it becomes increasingly challenging. Therefore, AI-based malware detection methods are essential to replace conventional fingerprint-based approaches that struggle to identify new types of malwares. In this research, we explore several stages: data collection, pre-processing, modeling, and evaluation. Our focus is on identifying optimal combinations of pre-processing and modeling methods to achieve high performance in Android malware detection. Specifically, we implement three feature selection methods, such as X<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> (Chi-Square), Information Gain, and ReliefF, in the pre-processing stage. Similarly, in the modeling stage, we employ three classification algorithms, namely: Random Forest, AdaBoost, and Gradient Boosting, to learn patterns from the data. Our most effective solution is the combination of the AdaBoost classifier with the ReliefF feature selection method, achieving 99.4% accuracy and F1-Score.