基于深度学习的恶意软件检测技术研究  

Research on Malware Detection Technology Based on Deep Learning

作　　者：王海宽[1] WANG Haikuan(Jincheng Vocational and Technical College,Jincheng Shanxi 048026,China)

机构地区：[1]晋城职业技术学院,山西晋城048026

出　　处：《佳木斯大学学报（自然科学版）》2023年第2期44-48,76,共6页Journal of Jiamusi University：Natural Science Edition

基　　金：晋城职业技术学院2022年校级课题(LX2216)。

摘　　要：近些年来,恶意软件正在以惊人的速度增长,并通过各式各样的混淆技术实现良好的伪装,从而给恶意软件的检测任务带来不小的挑战。此外,恶意软件种类及其变体多种多样,也给研究带来一定的困难。不依赖于程序执行的基于静态的检测方法可以快速有效地检测已知的恶意软件,但对未知的恶意软件检测效果不够理想。基于监控恶意程序执行过程的动态检测方法对未知和复杂的恶意软件表现良好但是性能不高。这些年来,基于深度学习的检测技术逐渐成熟,并随着计算机视觉方法不断发展,将恶意软件与可视化方法进行结合已经成为了一种新的趋势。对常见的恶意软件混淆技术进行简要的介绍,对常见检测方法进行总结,并对基于计算机视觉与深度学习的Windows恶意软件同源性检测方法进行研究。In recent years,malware is growing at an alarming rate,and through a variety of obfuscation technologies to achieve good camouflage,which brings no small challenge to the detection of malware.In addition,the variety of malware and its variants also bring some difficulties to the research.Static detection methods that do not rely on program execution can quickly and effectively detect known malware,but the detection effect for unknown malware is not ideal.The dynamic detection method based on monitoring the execution process of malicious programs performs well for unknown and complex malicious software,but its performance is not high.In recent years,the detection technology based on deep learning has gradually matured,and with the continuous development of computer vision methods,the combination of malware and visual methods has become a new trend.This paper briefly introduces common malware obfuscation technologies,summarizes common detection methods,and researches Windows malware homology detection methods based on computer vision and deep learning.

关 键 词：恶意软件 混淆技术 同源性分析 安全技术 

分 类 号：TP309[自动化与计算机技术—计算机系统结构] TP311[自动化与计算机技术—计算机科学与技术]