ReDroidDet: Android Malware Detection Based on Recurrent Neural Network
Mothanna Almahmoud, Dalia Alzu’bi, Qussai Yaseen
Abstract
Android still has the first rank in terms of market share in comparing to other operating systems. Due to its flexible publishing policy, companies are developing many applications in order to serve user needs. The official market of Android Google Play store is characterized by its support for the unofficial stores, and it does not impose many restrictions on developers during the publishing process. These features were a major reason for making it become the most vulnerable platform to cyber criminals, as users are suffering from the problems of exposure to malicious applications that breach their privacy or damage their devices. In this research, a novel model is devised based on a combination of four static features, namely; permissions, API calls, monitoring system events, and permission rate. Specifically, the dataset consists of 2,820 samples of both malware and benign applications. This paper proposes a new architecture of Recurrent Neural Network (RNN) that can perform the detection process better than traditional machine learning algorithms. The experimental results shown that the proposed model has scored 98.58 level of accuracy, and it has promising results in Android malware detection.