基于深度学习的对抗攻击发展研究  

Research on the development of adversarial attacks based on deep learning

作　　者：卢彦利 石雪莹 刘光晓 柳雪飞 文小慧 李章敏 蒋正锋[1,4] Lu Yanli;Shi Xueying;Liu Guangxiao;Liu Xuefei;Wen Xiaohui;Li Zhangmin;Jiang Zhengfeng(College of Mathematics,Physics and Electronic Information Engineering,Guangxi Minzu Normal University,Chongzuo 532200,China;School of Statistics and Data Science,Jiangxi University of Finance and Economics,Nanchang 330013,China;Xinfa Primary School,Chahe Town,Weining County,Bijie 553105,China;School of Computer Science,Wuhan University,Wuhan 430072,China)

机构地区：[1]广西民族师范学院数理与电子信息工程学院,崇左532200 [2]江西财经大学统计与数据科学学院,南昌330013 [3]威宁县岔河镇新发小学,毕节553105 [4]武汉大学计算机学院,武汉430072

出　　处：《现代计算机》2024年第8期44-49,共6页Modern Computer

基　　金：国家级大学生创新创业项目(202210604038)。

摘　　要：随着深度学习在各领域的广泛应用,对抗攻击问题引起学术界与工业界的关注。首先概述了对抗攻击的背景,包括对抗攻击的定义、分类以及与传统的机器学习安全问题的区别。然后讨论了对抗样本生成及攻击策略,以及白盒攻击和黑盒攻击等攻击手段。最后总结了对抗攻击的意义,并展望未来研究方向,期待通过研究和探索提高深度学习模型的安全性和可靠性。With the widespread application of deep learning in various fields,the issue of adversarial attacks has attracted at⁃tention from both academia and industry.Firstly,the background of adversarial attacks is outlined,including the definition,classifi⁃cation,and differences from traditional machine learning security issues.Then we discussed adversarial sample generation and at⁃tack strategies,as well as attack methods such as white box and black box attacks.Finally,the significance of adversarial attacks was summarized,and future research directions were looked forward to improving the security and reliability of deep learning mod⁃els through research and exploration.

关 键 词：深度学习 对抗攻击 数据攻击 模型攻击 防御策略 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]