DeepDetector: Android Malware Detection using Deep Neural Network
Dongfang Li, Zhaoguo Wang, Yibo Xue · 2018
The security issue of Android Smart Phones has been most concerned by the users these years. We propose a novel method to extract exhaustive features from Android applications and use the deep-learning-based method to detect malicious applications. Then we implement an automatic detection engine, DeepDetector, to detect malicious applications. Furthermore, the detection model can identify the fine-grained malware family at the same time. We conducted the evaluation and analysis with thousands of malicious applications and benign applications from a public dataset. The results show that DeepDetector can detect 97% of the malware at 0.1 % false positive rate (FPR) and achieves a 96% precision when showing the detailed malware families. Besides, the relationship between the detection performance and the architecture of the neural network is also discussed in our work.