基于集成学习的安卓恶意软件特征提取与检测方法

打开文本图片集

中图分类号：TP311.5 文献标志码：A 文章编号：2095-2945（2025）18-0045-05

Abstract：Android，asthemostpopularoperatingsystemtodayofersconvenience tousersthroughitsopennessandwide application.However，thissameopennessalsoprovidesopportunitiesformalwaredevelopment，posingsignificantthreatstousers personalprivacyanddatasecurity.Toaddressthisisue，thisstudyproposesanintegratedlearning-basedmethodforfeature extractionanddetectionofAndroidmalware.TheauthorizationrequestofAndroidAPKisextractedasfeaturepointsthrough automatedscripts，combinedwithanenhancedsupportvector machine（E-SVM）modelandaconvolutionalneuralnetwork （CNN）modelforintegratedlearningtraining，generatedahybridmodel，andusedtoimprovethedetectionrateofAndroid malware.Final experimental data shows that the detection accuracy rate for malware reaches more than 96%

Keywords：malware；machinelearning;deep learning;inheritancelearning;feature extractionand detection

现如今，Android作为全世界最为流行的作业系统，占有着全世界百分之八十的市场份额，已经拥有了成千上万的用户。（剩余5270字）