MACHINE LEARNING BASED ENSEMBLE CLASSIFIER FOR ANDROID MALWARE DETECTION

Abstract

Malware problem has infiltrated into every aspect of cyber space including Android mobiles. Due to proliferation of Android applications and widespread usage of smartphones, malware problem is causing significant damage to mobile users and application vendors. With the emergence of Artificial Intelligence (AI), machine learning (ML) models are widely used for detection of Android malware. However, many of the existing methods focused on static or dynamic data to train classifiers for malware detection. In this paper, we propose an ensemble model with intelligent methods that are empirically selected. Only the malware detection models with highest accuracy are chosen to be part of stacking ensemble model. An algorithm named Stacking Ensemble for Automatic Android Malware Detection (SE-AAMD)is proposed and implemented. We made three experiments with the same algorithm but three different datasets reflecting features obtained through different modus operandi. Each dataset is found to have influence on the performance of the models. However, in all experiments, the ensemble approach showed highest performance. The proposed method can be used in improving security for Android devices and applications.