TENSORSHIELD: Tensor-based Defense Against Adversarial Attacks on Images

被引:0
作者
Entezari, Negin [1 ]
Papalexakis, Evangelos E. [1 ]
机构
[1] Univ Calif Riverside, Riverside, CA 92521 USA
来源
2022 IEEE MILITARY COMMUNICATIONS CONFERENCE (MILCOM) | 2022年
关键词
adversarial machine learning; deep neural networks; image classification;
D O I
10.1109/MILCOM55135.2022.10017763
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Recent studies have demonstrated that machine learning approaches like deep neural networks (DNNs) are easily fooled by adversarial attacks. Subtle and imperceptible perturbations of the data are able to change the result of deep neural networks. Leveraging vulnerable machine learning methods raises many concerns, especially in domains where security is an important factor. Therefore, it is crucial to design defense mechanisms against adversarial attacks. For the task of image classification, unnoticeable perturbations mostly occur in the high-frequency spectrum of the image. In this paper, we utilize tensor decomposition techniques as a preprocessing step to find a low-rank approximation of images that can significantly discard high-frequency perturbations. Recently a defense framework called SHIELD [1] could "vaccinate" Convolutional Neural Networks (CNN) against adversarial examples by performing random-quality JPEG compressions on local patches of images on the ImageNet dataset. Our tensor-based defense mechanism outperforms the SLQ method from SHIELD by 14% against Fast Gradient Descent (FGSM) adversarial attacks, while maintaining comparable speed.
引用
收藏
页数:6
相关论文
共 35 条
  • [1] Abadi M, 2016, PROCEEDINGS OF OSDI'16: 12TH USENIX SYMPOSIUM ON OPERATING SYSTEMS DESIGN AND IMPLEMENTATION, P265
  • [2] Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
    Akhtar, Naveed
    Mian, Ajmal
    [J]. IEEE ACCESS, 2018, 6 : 14410 - 14430
  • [3] Baldi P., 2012, P ICML WORKSH UNS TR, P37
  • [4] Bhagoji A. N., 2017, ARXIV
  • [5] Cornelius C, 2019, Arxiv, DOI arXiv:1902.00541
  • [6] SHIELD: Fast, Practical Defense and Vaccination for Deep Learning using JPEG Compression
    Das, Nilaksh
    Shanbhogue, Madhuri
    Chen, Shang-Tse
    Hohman, Fred
    Li, Siwei
    Chen, Li
    Kounavis, Michael E.
    Chau, Duen Horng
    [J]. KDD'18: PROCEEDINGS OF THE 24TH ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY & DATA MINING, 2018, : 196 - 204
  • [7] Deng J, 2009, PROC CVPR IEEE, P248, DOI 10.1109/CVPRW.2009.5206848
  • [8] Doersch C, 2021, Arxiv, DOI arXiv:1606.05908
  • [9] All You Need Is Low (Rank): Defending Against Adversarial Attacks on Graphs
    Entezari, Negin
    Al-Sayouri, Saba A.
    Darvishzadeh, Amirali
    Papalexakis, Evangelos E.
    [J]. PROCEEDINGS OF THE 13TH INTERNATIONAL CONFERENCE ON WEB SEARCH AND DATA MINING (WSDM '20), 2020, : 169 - 177
  • [10] Guo C., 2017, arXiv