Identifying Latent Android Malware from Application’s Description using LSTM

Abstract

Android is the most popular mobile system in the world where many applications provide convenient and diverse functions on top of it for our daily lives. However, a new Android malware is revealed every 10 seconds and the official application markets still consists of malicious and undetected applications due to the limitation of the existing malware detection techniques. In this paper, we propose an approach to identify the latent Android malware from application’s description using Long Short-Term Memory (LSTM) technique. The actual permissions requested by source code and permissions predicted from the description using semantics analysis to are compared to verify the consistency. If an application requests a permission undeclared in the description or homogeneous applications, it will be considered as a latent Android malware. This approach can surely provide more secure environment for the end-users before they install the applications.