安卓恶意应用检测中的特征研究与应用

In recent years, Android platform has a rapid development because of its superior customer experience and openness. Meanwhile the platform has become the primary target of malicious attackers. It has become the major challenge of Android market management and malicious application detection that the category of application from Android platform tend to be diverse and the behavior of malicious application tend to be complex. For the continued deterioration of the Android security situation, this paper analysis the available features in malicious application detection and some related datasets. The work can provide effective support for subsequent follow-up study about the features in malicious applications detection. So the research of automatic classification of Android applications is of great significance. We use two machine learning algorithms to detect malicious applications with 116,028 Android applications and 7 kinds of static features. After a great deal of experiments, our detection accuracy rate reached 99.38% with FPR 0.03%.