A Systematical Study for Deep Learning Based Android Malware Detection

Nowadays with the development of smartphone and its operating system, such as Android, the amount of mobile malware is correspondingly increasing. To protect the privacy security of both users and manufacturers, Android malware detection has received a lot of research focuses. Traditional methods mainly rely on static analysis or dynamic monitoring, which is either software sensitive or time-consuming. Recently, with the development of machine learning and deep learning techniques, many efforts introduced such learning-based techniques into the Android malware detection and achieved promising detection results as well as substantially reduced time costing. Nevertheless, there is still a lack of a comprehensive summary at the deep learning technical level for these learning-based malware detection works. As a result, it is limited to improve the detection technique referring to existing works from a global map. To address the challenge, in this paper, we systematically study existing deep learning-based malware detection works and classify them from the technique perspective. We also conclude the advantages and threats within each category of detection technique and provide a concrete technical reference for future improvement work.