摘要
生成对抗网络是一种通过对抗过程来学习原始样本分布规律的深度学习技术。该方法非常适合于存在对抗行为的智能处理工作,如语音翻译、人脸识别、图像分类、恶意软件样本生成等领域。文章综述生成对抗网络方法应用于恶意软件生成领域所取得的研究成果,对所涉及的特征提取技术、分类检测算法、生成对抗网络原理及对抗样本生成模型等方面的研究成果进行深入的分析和比较,同时对将基于GAN的生成对抗样本技术应用于网络安全领域进行展望,最后提出生成对抗样本技术所面临的挑战。
Generative adversarial networks is a deep learning technique that learns the distribution of original samples by adversarial processes.This method is very suitable for intelligent processing work with adversarial behavior,such as speech translation,face recognition,image classification,malware sample generation and other fields.This paper reviews the research results of GAN model applied in the field of malicious software generation,and deeply analyses and compares the research results of feature extraction technology,classification detection algorithm,principle of GAN and adversarial sample generation model.At the same time,the application of adversarial sample generation technology on GAN to the field of network security is prospected.Finally,the challenges are presented.
作者
王树伟
周刚
巨星海
陈靖元
WANG Shuwei;ZHOU Gang;JU Xinghai;CHEN Jingyuan(Information Engineering University, Zhengzhou 450001, China)
出处
《信息工程大学学报》
2019年第5期616-621,共6页
Journal of Information Engineering University
关键词
生成对抗网络
机器学习
特征提取
恶意代码检测
对抗样本生成
generative adversarial networks
machine learning
feature extraction
malware detection
adversarial sample generation