A multi-level feature extraction technique to detect moble botnet

Android malware detection has been heavily studied, which classical android malware detecting approaches are signature-based or behavior-based detection based on the files itself, but little attention has been directed to the network traffics generated by android malwares, known as mobile botnet. In this paper, a multi-level feature extraction technique was presented to detect the mobile botnet. It means that it's not only extract features from the TCP/IP basic info level, but also extract info from traffic level, more over from the content level such as http content, rather than using more features at the single level. Finally, these different feature sets were combined into one feature set which is used by the classifiers for training/testing. Our method is compared against other Android malware code detection and found more efficient features in mobile botnet detection.