基于特征贡献度的安卓恶意应用检测  

Android malicious application detection based on feature contribution degree

作　　者：刘启川 覃仁超[1] 刘玲 卜得庆 袁平[1] LIU Qi-chuan;QIN Ren-chao;LIU Ling;BU De-qing;YUAN Ping(School of Computer Science and Technology,Southwest University of Science and Technology,Mianyang 621010,China)

机构地区：[1]西南科技大学计算机科学与技术学院,四川绵阳621010

出　　处：《计算机工程与设计》2020年第4期928-932,共5页Computer Engineering and Design

基　　金：西南科技大学博士基金项目(10zx7154)。

摘　　要：为提高Android恶意软件检测准确率,提出一种基于特征贡献度的特征选择算法。针对现有Android应用数据集特征的分布特点,通过计算特征的类内以及类间贡献度,设定阈值筛选出贡献度高的特征数据,用于恶意应用检测分类。实验结果表明,所提算法能有效且可靠地检测恶意应用,其准确率和召回率十分接近,适用于恶意应用检测;与传统特征选择算法相比,该算法可以在较少特征数量的情况下达到理想的检测效果。To improve the detection accuracy of Android malware,a feature selection algorithm based on feature contribution was proposed.Aiming at the distribution characteristics of the existing Android application dataset features,the intra-class and inter-class contribution degree of features were calculated and the threshold value was set to filter out the feature data with high contribution degree,for malicious application detection classification.The experimental results show that,the proposed algorithm can detect malicious applications effectively and reliably,and its accuracy and recall rate are very close,which is suitable for malicious application detection.Compared with the traditional feature selection algorithm,the algorithm can achieve the ideal detection effects with fewer feature numbers.

关 键 词：ANDROID系统 恶意应用 特征选择算法 特征贡献度 机器学习 

分 类 号：TP393.0[自动化与计算机技术—计算机应用技术]