Deep learning in detection of mobile malware

Deep learning in detection of mobile malware

연구 분야: Safety

학회: Journal of Computing Sciences in Colleges, Volume 36, Issue 3

Traditional malware detection techniques have been widely adopted in desktop and laptop computers. In this paper, we investigate how deep learning models can be adapted to detecting malware in mobile devices. We use a deep neural network (DNN) implementation of deep learning called DeepLearning4J (DL4J) to generate our models for mobile malware detection. The models detect mobile malware with accuracy rates ranging between 97% and 99% when applied to two types of malicious datasets. For our DNN models we vary the layers to determine effectiveness of the models in detecting mobile malware. Further analysis reveals that DNN models continue to improve in accuracy of detecting mobile malware as more layers are added to the models. This study demonstrates the practicability of using the DNN to continually learn from the past malware attacks and finally be able to predict new types of attacks. This paper sheds light on a new direction of examining malware prevention, detection and prediction and motivates future direction of exploring new strains of mobile malware that can be detected using machine learning.