计算机与数字工程2023，Vol.51Issue(10)：2346-2350,5.DOI:10.3969/j.issn.1672-9722.2023.10.025
基于深度学习的可扩展Android恶意软件检测和分类方案
Scalable Android Malware Detection and Classification Scheme Based on Deep Learning
摘要
Abstract
The Android operating system is currently one of the mainstream operating systems in mobile devices.It has a large user group,so many malicious Android software have also appeared.Every year,researchers will propose some new Android mal-ware analysis frameworks to defend against real-world Android malware applications.This article uses mainstream deep learning al-gorithms,builds a suitable neural network,and adds rectified linear units to realize Android malware detection and classification.Through training the network,this paper finally gets a relatively good malicious detector(binary classifier)and three multi-class classifiers results-based on the static malware binary classifier with 95.74%accuracy and multi-class classifier with 92.98%accura-cy.The dynamic-based malware category multi-class classifier has an accuracy of 84.48%and the dynamic-based malware family multi-class classification has an accuracy of 60.34%.关键词
Android/恶意软件/深度学习/神经网络/线性修正单元/二分类器/多分类器Key words
Android/malware/deep learning/neural network/rectified linear unit/binary classifier/multi-class classifier分类
信息技术与安全科学引用本文复制引用
毛慈伟,刘万里,李荣臻,尹魏昕..基于深度学习的可扩展Android恶意软件检测和分类方案[J].计算机与数字工程,2023,51(10):2346-2350,5.基金项目
国家自然科学基金项目(编号:61973161,61991404) （编号:61973161,61991404）
江苏省科技计划项目(编号:BE2021610) （编号:BE2021610）
江苏省教育厅未来网络科研基金项目(编号:FNSRFP-2021-YB-05,FNSRFP-2021-ZD-4)资助. （编号:FNSRFP-2021-YB-05,FNSRFP-2021-ZD-4）
