Unlearning Backdoor Attacks through Gradient-Based Model Pruning

被引:0
|
作者
Dunnett, Kealan [1 ,2 ]
Arablouei, Reza [2 ]
Miller, Dimity [1 ]
Dedeoglu, Volkan [1 ,2 ]
Jurdak, Raja [1 ]
机构
[1] Queensland Univ Technol, Brisbane, Qld, Australia
[2] CSIROs Data61, Canberra, ACT, Australia
关键词
backdoor attack; backdoor mitigation; model pruning; unlearning;
D O I
10.1109/DSN-W60302.2024.00021
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In the era of increasing concerns over cybersecurity threats, defending against backdoor attacks is paramount in ensuring the integrity and reliability of machine learning models. However, many existing approaches require substantial amounts of data for effective mitigation, posing significant challenges in practical deployment. To address this, we propose a novel approach to counter backdoor attacks by treating their mitigation as an unlearning task. We tackle this challenge through a targeted model pruning strategy, leveraging unlearning loss gradients to identify and eliminate backdoor elements within the model. Built on solid theoretical insights, our approach offers simplicity and effectiveness, rendering it well-suited for scenarios with limited data availability. Our methodology includes formulating a suitable unlearning loss and devising a model-pruning technique tailored for convolutional neural networks. Comprehensive evaluations demonstrate the efficacy of our proposed approach compared to state-of-the-art approaches, particularly in realistic data settings.
引用
收藏
页码:46 / 54
页数:9
相关论文
共 50 条
  • [31] Model-reduced gradient-based history matching
    Małgorzata P. Kaleta
    Remus G. Hanea
    Arnold W. Heemink
    Jan-Dirk Jansen
    Computational Geosciences, 2011, 15 : 135 - 153
  • [32] Gradient-Based Supported Model Computation in Vector Spaces
    Takemura, Akihiro
    Inoue, Katsumi
    LOGIC PROGRAMMING AND NONMONOTONIC REASONING, LPNMR 2022, 2022, 13416 : 336 - 349
  • [33] Model Debiasing via Gradient-based Explanation on Representation
    Zhang, Jindi
    Wang, Luning
    Su, Dan
    Huang, Yongxiang
    Cao, Caleb Chen
    Chen, Lei
    PROCEEDINGS OF THE 2023 AAAI/ACM CONFERENCE ON AI, ETHICS, AND SOCIETY, AIES 2023, 2023, : 193 - 204
  • [34] A Gradient-based reinforcement learning model of market equilibration
    He, Zhongzhi
    JOURNAL OF ECONOMIC DYNAMICS & CONTROL, 2023, 152
  • [35] Model-reduced gradient-based history matching
    Kaleta, Malgorzata P.
    Hanea, Remus G.
    Heemink, Arnold W.
    Jansen, Jan-Dirk
    COMPUTATIONAL GEOSCIENCES, 2011, 15 (01) : 135 - 153
  • [36] A Gradient-Based Constitutive Model for Shape Memory Alloys
    Tabesh M.
    Boyd J.
    Lagoudas D.
    Boyd, James (jgboyd@tamu.edu), 1600, Springer (03): : 84 - 108
  • [37] Gradient-based adaptation of continuous dynamic model structures
    La Cava, William G.
    Danai, Kourosh
    INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE, 2016, 47 (01) : 249 - 263
  • [38] A Gradient Control Method for Backdoor Attacks on Parameter-Efficient Tuning
    Gu, Naibin
    Fu, Peng
    Liu, Xiyu
    Liu, Zhengxiao
    Lin, Zheng
    Wang, Weiping
    PROCEEDINGS OF THE 61ST ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, ACL 2023, VOL 1, 2023, : 3508 - 3520
  • [39] Backdoor Defence for Voice Print Recognition Model Based on Speech Enhancement and Weight Pruning
    Zhu, Jiawei
    Chen, Lin
    Xu, Dongwei
    Zhao, Wenhong
    IEEE Access, 2022, 10 : 114016 - 114023
  • [40] Detecting Backdoor Attacks on Deep Neural Networks Based on Model Parameters Analysis
    Ma, Mingyuan
    Li, Hu
    Kuang, Xiaohui
    2022 IEEE 34TH INTERNATIONAL CONFERENCE ON TOOLS WITH ARTIFICIAL INTELLIGENCE, ICTAI, 2022, : 630 - 637