An effective mobile malware detection framework for android security

摘要

>Mobile malware is considered as one of the crucial security challenges due to its high volume and quick variety, especially on the Android platform. Many researches have been proposed to detect malware, but some of them suffer low detection accuracy or high time consumption. This research implements an effective mobile malware detection framework by proposing a new feature selection method, which is term frequency-sample frequency differentiation (TF-SFD), to reduce the features with little importance. In addition, a false positive rate (FPR) filter is proposed based on sample frequency differentiation (SFD) for reducing FPR. We investigate four machine learning methods and the experimental results show that the TF-SFD combining with random forest (RF) classifier performs best in terms of accuracy in detecting malware on Android, which obtains 92.54% testing accuracy.