摘要
为了提高Android恶意软件检测的准确率和效率,提出一种在静态分析技术基础上利用自动编码器(AE)网络和深度信念网络(DBN)结合的Android恶意软件检测方案。首先通过静态分析技术,提取了权限、动作、组件和敏感APIs作为特征信息,其次通过AE对特征数据集进行降维,最后结合DBN进行更深层次的特征抽象学习,并训练DBN来进行恶意代码检测。实验结果证明,提出的方案与DBN,SVM和KNN进行比较,提高了检测效率和准确率,降低了误报率。
In order to improve the accuracy and efficiency of Android malware detection,an Android malware detection scheme is proposed based on static analysis technology using Auto Encoder( AE)network and Deep belief network( DBN). Firstly,statistical analysis technology is used to extract the permissions,actions,components and sensitive APIs as the features. Secondly,AE is used to reduce dimensions of the features dataset. Finally,the scheme combined with DBN to continue furture learning of abstract feature and trained DBN to conduct the malicious. The experimental results demonstrate that this method is compared with DBN,SVM,and KNN,which improves detection efficiency and accuracy. Besides,the false positive rate also is reduced.
作者
吴招娣
徐洋
谢晓尧
WU Zhaodi;XU Yang;XIE Xiaoyao(Key Laboratory of Information and Computing Science of Guizhou Province, Guizhou Normal University, Guiyang, Guizhou 550001, China)
出处
《贵州师范大学学报（自然科学版）》
CAS
2019年第3期96-101,共6页
Journal of Guizhou Normal University：Natural Sciences
基金
中央引导地方科技发展专项资金项目(黔科中引地〔2018〕4008)
贵州省科技合作计划重点项目(黔科合LH字[2015]7763)
