Detecting Android Malicious Applications using Dynamic Malware Analysis and Machine Learning

Abstract

With the rise in usage of smartphones, the number of malicious apps targeting the Android mobile platform has risen dramatically. These days, malware is coded so carefully that it is extremely difficult to recognize. Traditional malware detection methods are outdated because current malware uses sophisticated obfuscation techniques to hide its functionalities from scanning engines. This paper presents an approach based on dynamic malware analysis for the identification of malicious samples. In this, the applications are executed in a virtual environment (Sandbox) to determine the behavior of an application. The proposed model is evaluated on 3547 apps. The results illustrate that the proposed approach is found to be more accurate and effective for the identification of Android malware. The accuracy acquired by the proposed model is 98.26%.