Fest: A feature extraction and selection tool for Android malware detection

摘要: Android has become one of the most popular mobile operating systems because numerous applications (apps) it provides. However, malware downloaded from third-party markets threatens users' privacy, and them remain undetected lack efficient accurate detecting techniques. Prior efforts on detection attempted to build precise classification models by manually choosing features, few used any feature selection algorithms help pick typical features. In this paper, we present Feature Extraction Selection Tool (Fest), a feature-based machine learning approach for detection. We first implement extraction tool, AppExtractor, which is designed extract such as permissions or APIs, according predefined rules. Then propose algorithm, FrequenSel. Unlike existing features calculating their importance, FrequenSel selects finding difference frequencies between benign apps, are frequently in rarely apps more important distinguish apps. experiments, evaluate our with 7972 results show that Fest gets nearly 98% accuracy recall, only 2% false alarms. Moreover, takes 6.5s analyze an app common PC, very time-efficient markets.