Android malware detection using machine learning

Abstract

Because Android is open-source and widely used, it has emerged as the most popular mobile operating system. But because of its widespread use, fraudsters who disseminate malicious software have found it to be a prime target. Although they work well for known threats, traditional signature-based malware detection techniques miss novel or unidentified variations, increasing the danger of zero-day assaults. This paper suggests a machine-learning-based detection framework improved using Genetic Algorithm (GA) for optimal feature selection in order to get over these restrictions. By selecting the most discriminative and pertinent characteristics from big feature sets, the GA lowers dimensionality without sacrificing accuracy. Machine learning classifiers like Support Vector Machines (SVM) and Neural Networks (NN) are then trained using these improved features. According to experimental data, the suggested method reduces computing complexity by almost half while achieving detection accuracy of over 92.56%. This study shows a scalable, effective, and lightweight malware detection system.