基于组合机器学习算法的Android恶意软件检测  

Android malware detection based on combinatorial machine learning algorithm

作　　者：丘惠军 连耿雄 刘则君 QIU Hui-jun;LIAN Geng-xiong;LIU Ze-jun(Limited Company of Shenzhen Power Supply Bureau,Shenzhen 518020,Guangdong Province,China;China Electric Power Operation (Beijing) Information Technology Co.,Ltd.,Beijing 100089,China)

机构地区：[1]深圳供电局有限公司,广东深圳518020 [2]中电运行(北京)信息技术有限公司,北京100089

出　　处：《信息技术》2019年第7期59-64,共6页Information Technology

基　　金：中国南方电网科技项目(090000KK52170049)

摘　　要：Android系统的开源特性使得应用市场缺乏完善的监管,严重影响移动用户的信息安全。然而,现有Android恶意软件检测方法主要依赖于恶意软件签名库,难以应对无签名的未知软件。为此,文中提出一种基于组合机器学习算法的Android恶意软件检测方法,该方法充分利用Android应用的多项属性特征,使用多个策略组合分析不同类型机器学习分类器的决策结果。实验结果表明,文中方法能够有效提高Android恶意软件检测的有效性和准确性。Android system nature of open source makes the market lack comprehensive supervision,which seriously affects information security of mobile users. However,the existing Android malware detection approach relies heavily on the malware signature library,which makes it difficult to effectively respond to unsigned,unknown malware. Therefore,an Android malware detection method based on combinatorial machine learning algorithm is proposed,which makes full use of the multiple features of Android application and uses multiple strategies to combine the decision results of different types of machine learning classifiers. The experimental results show that it can effectively improve the effectiveness and accuracy of Android malware detection.

关 键 词：Android移动应用 恶意软件检测 机器学习 静态特征提取 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]