Comprehensive Behavior Profiling for Proactive Android Malware Detection

摘要: We present a new method of screening for malicious Android applications that uses two types information about the application: permissions application requests in its installation manifest and metric called percentage valid call sites (PVCS). PVCS measures riskiness based on data flow graph. The is used with machine learning algorithms to classify previously unseen as or benign high degree accuracy. Our classifier outperforms previous state art by significant margin, particularly low false positive rates. Furthermore, evaluation performed malware families were not training phase, simulating accuracy yet be developed. found our SEND_SMS permission are specific pieces most useful classifier.