Android Malware Detection Methods Based on Convolutional Neural Network: A Survey

Android malware detection(AMD) is a challenging task requiring many factors to be considered during detection, such as feature extraction and processing, performance evaluation, and many available datasets. AMD aims to develop more effective algorithms and models to protect users' privacy and data security. Deep learning(DL) has recently received considerable attention in AMD, especially convolutional neural networks(CNN), which can handle binary data in Android applications more efficiently, avoid feature engineering, and cope well with rapid malware updates. However, CNN-based AMD papers are increasing and scattered. This article tries to review AMD techniques based on CNN. First, the main steps in AMD are systematically reviewed, such as data collection and preprocessing, feature extraction, feature representation, model training, and model evaluation. Then, the literature is summarized according to the different features used in detection. Finally, the challenges of AMD and future research directions are presented.