Abstract: Android is the most popular operating system by far, which has the highest market share. Malicious software based on Android platform also presents explosive growth, but currently there are no effective means, which can detect the Android malicious behavior. In this paper, through analyzing the characteristics of the Android malicious behavior, it uses the machine learning algorithm based on Bayesian networks to detect the Android malicious behavior. Beyond that, this paper extracts the static characteristics of the Android file based on the static analysis method, which has realized the combination of static analysis and the Bayesian network. In the end, through the experiment, it verifies the effectiveness of the Android malicious behavior detection model.

Key words: Android, machine learning, feature selection, Bayesian network

摘要： Android操作系统是市场占有率最高的移动操作系统，基于Android平台的恶意软件也呈现爆发式的增长，而目前仍然没有有效的手段进行Android恶意行为的检测，通过分析Android恶意行为的特点，采用基于贝叶斯网络的机器学习算法进行Android恶意行为的检测，通过静态分析的方法进行Android文件静态特征的提取，将Android恶意应用的静态分析与贝叶斯网络相结合，最后通过使用提出的方法构建贝叶斯网络模型，通过实验验证了提出的Android恶意行为检测模型的有效性。

关键词: Android, 机器学习, 特征选择, 贝叶斯网络