基于多特征的Android移动设备恶意代码检测

Multi-features-based malware code detection for Android mobile devices

目前基于机器学习的Android移动应用恶意代码检测技术侧重于对单一特征进行检测分析,不能充分利用多类特征对Android恶意代码检测所起的作用、不能充分使用不同机器学习算法对某类行为特征的最优算法。文中采用了基于行为的静态检测技术,通过数据挖掘算法选取三类特征,使用多特征融合模型来识别Android恶意应用,准确率更高、误报率更低。

At present,malware code detection for Android mobile application based on machine learning focuses on single feature. While single feature could not make the best of multi-class features in Android malware code detection. The paper adopted the static detection technology based on behavior,handled three classes of features through the data mining algorithm and used multi-class features fusion model to identify unknown Android malware with higher accuracy and lower FP rate.