Android Malware Detection Using Machine Learning with Feature Selection Based on The Genetic Algorithm

The considerable growth in the number of Android devices has made this platform the primary target for malware and has resulted in the demand for robust detection measures to safeguard user data and system integrity. This study presents an innovative Android malware detection architecture that combines machine learning and feature selection through Genetic Algorithms (GA). Although machine learning approaches are good at identifying malicious behavior, the performance of the underlying model is reliant on the features chosen. Because of this, GA an optimization algorithm that mimics biological evolution is used in this study to identify the most relevant features of Android applications to minimize dimensionality and improve accuracy. The hybrid approach proposed in this framework, incorporates both static and dynamic features of Android applications including, permissions, API calls, and network behavior; the framework then employs GA to refine the feature set applied to machine learning algorithms including Random Forest, Support Vector Machine, and Neural Networks, to classify the applications. The practical and experimental findings demonstrate that GA-based feature selection significantly improves malware detection accuracy, precision, recall, and F1 score, while also reducing computational cost, and is therefore applicable in resource constrained settings.