A Deep Learning Approach to Android Malware Feature Learning and Detection

Abstract

The growing amount and diversity of Android malware has significantly weakened the effectiveness of the conventional defense mechanisms, and thus Android platform often remains unprotected from new and unknown malware. To address these limitations, we propose DroidDeep, a malware detection approach for the Android platform based on the deep learning model. Deep learning emerges as a new area of machine learning research that has attracted increasing attention in artificial intelligence. To implement this, we first extract five types of features from the static analysis of Android apps. Then, we build the deep learning model to learn features from Android apps. Finally, the learned features are used to detect unknown Android malware. In an experiment with 3,986 benign apps and 3,986 malware, DroidDeep outperforms several existing malware detection approaches and achieves a 99.4% detection accuracy. Moreover, DroidDeep can achieve a remarkable run-time efficiency which makes it very easy to adapt to a lager scale of real-world Android malware detection.