基于集成學(xué)習(xí)的安卓惡意軟件特征提取與檢測(cè)方法

打開(kāi)文本圖片集

中圖分類(lèi)號(hào)：TP311.5 文獻(xiàn)標(biāo)志碼：A 文章編號(hào)：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

現(xiàn)如今，Android作為全世界最為流行的作業(yè)系統(tǒng)，占有著全世界百分之八十的市場(chǎng)份額，已經(jīng)擁有了成千上萬(wàn)的用戶(hù)。（剩余5270字）