《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (9): 2732-2741.DOI: 10.11772/j.issn.1001-9081.2021071339

• 网络空间安全 • 上一篇    下一篇

图像分类中的白盒对抗攻击技术综述

魏佳璇1(), 杜世康1, 于志轩1,2, 张瑞生1   

  1. 1.兰州大学 信息科学与工程学院,兰州 730000
    2.兰州大学第一医院,兰州 730000
  • 收稿日期:2021-07-26 修回日期:2021-09-29 接受日期:2021-10-08 发布日期:2021-10-25 出版日期:2022-09-10
  • 通讯作者: 魏佳璇
  • 作者简介:杜世康(1997—),男,甘肃武威人,硕士研究生,主要研究方向:对抗机器学习;
    于志轩(1983—),男,甘肃兰州人,博士研究生,主要研究方向:机器学习、深度学习;
    张瑞生(1962—),男,甘肃兰州人,教授,博士,主要研究方向:可解释机器学习、复杂网络分析、图像识别与分析、服务计算、化学信息学、生物信息学。
  • 基金资助:
    甘肃省自然科学基金资助项目(20YF8FA080)

Review of white-box adversarial attack technologies in image classification

Jiaxuan WEI1(), Shikang DU1, Zhixuan YU1,2, Ruisheng ZHANG1   

  1. 1.School of Information Science and Engineering,Lanzhou University,Lanzhou Gansu 730000,China
    2.The First Hospital of Lanzhou University,Lanzhou Gansu 730000,China
  • Received:2021-07-26 Revised:2021-09-29 Accepted:2021-10-08 Online:2021-10-25 Published:2022-09-10
  • Contact: Jiaxuan WEI
  • About author:DU Shikang, born in 1997, M. S. candidate. His research interests include adversarial machine learning.
    YU Zhixuan, born in 1983, Ph. D. candidate. His research interests include machine learning, deep learning.
    ZHANG Ruisheng, born in 1962, Ph. D., professor. His research interests include interpretable machine learning, complex network analysis, image recognition and analysis, service computing, chemoinformatics, bioinformatics.
  • Supported by:
    Natural Science Foundation of Gansu Province(20YF8FA080)

摘要:

在深度学习中图像分类任务研究里发现,对抗攻击现象给深度学习模型的安全应用带来了严峻挑战,引发了研究人员的广泛关注。首先,围绕深度学习中用于生成对抗扰动的对抗攻击技术,对图像分类任务中重要的白盒对抗攻击算法进行了详细介绍,同时分析了各个攻击算法的优缺点;然后,分别从移动终端、人脸识别和自动驾驶三个现实中的应用场景出发,介绍了白盒对抗攻击技术的应用现状;此外,选择了一些典型的白盒对抗攻击算法针对不同的目标模型进行了对比实验并分析了实验结果;最后,对白盒对抗攻击技术进行了总结,并展望了其有价值的研究方向。

关键词: 对抗样本, 白盒对抗攻击, 深度学习, 图像分类, 人工智能安全

Abstract:

In the research of image classification tasks in deep learning, the phenomenon of adversarial attacks brings severe challenges to the secure application of deep learning models, which arouses widespread attention of researchers. Firstly, around the adversarial attack technologies for generating the adversarial perturbations, the important white-box adversarial attack algorithms in the image classification tasks were introduced in detail, and the advantages and disadvantages of different attack algorithms were analyzed. Then, from three realistic application scenarios: mobile application, face recognition and autonomous driving, the application status of the white-box adversarial attack technologies was illustrated. Additionally, some typical white-box adversarial attack algorithms were selected to perform experiments on different target models, and the experimental results were analyzed. Finally, the white-box adversarial attack technologies were summarized, and their valuable research directions were prospected.

Key words: adversarial example, white-box adversarial attack, deep learning, image classification, artificial intelligence security

中图分类号: