Bypassing Backdoor Detection Algorithms in Deep Learning

被引:72
|
作者
Tan, Te Juin Lester [1 ]
Shokri, Reza [1 ]
机构
[1] Natl Univ Singapore NUS, Dept Comp Sci, Singapore, Singapore
关键词
D O I
10.1109/EuroSP48549.2020.00019
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Deep learning models are vulnerable to various adversarial manipulations of their training data, parameters, and input sample. In particular, an adversary can modify the training data and model parameters to embed backdoors into the model, so the model behaves according to the adversary's objective if the input contains the backdoor features, referred to as the backdoor trigger (e.g., a stamp on an image). The poisoned model's behavior on clean data, however, remains unchanged. Many detection algorithms are designed to detect backdoors on input samples or model parameters, through the statistical difference between the latent representations of adversarial and clean input samples in the poisoned model. In this paper, we design an adversarial backdoor embedding algorithm that can bypass the existing detection algorithms including the state-of-the-art techniques. We design an adaptive adversarial training algorithm that optimizes the original loss function of the model, and also maximizes the indistinguishability of the hidden representations of poisoned data and clean data. This work calls for designing adversary-aware defense mechanisms for backdoor detection.
引用
收藏
页码:175 / 183
页数:9
相关论文
共 50 条
  • [21] Computational Comparison of Deep Learning Algorithms for Object Detection
    Balafas, Vasileios
    Ploskas, Nikolaos
    25TH PAN-HELLENIC CONFERENCE ON INFORMATICS WITH INTERNATIONAL PARTICIPATION (PCI2021), 2021, : 79 - 83
  • [22] Object Detection Algorithms Based on Deep Learning and Transformer
    Fu, Miaomiao
    Deng, Miaolei
    Zhang, Dexian
    Computer Engineering and Applications, 2023, 59 (01): : 37 - 48
  • [23] A comparison framework for deep learning RFI detection algorithms
    Du Toit, Charl D.
    Grobler, Trienko L.
    Ludick, Danie J.
    MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY, 2024, 530 (01) : 613 - 629
  • [24] Understanding Deep Learning Algorithms for Object Detection and Recognition
    Suriya, S.
    Rajasekar, Rajesh Harinarayanan
    Shalinie, S. Mercy
    2019 11TH INTERNATIONAL CONFERENCE ON ADVANCED COMPUTING (ICOAC 2019), 2019, : 79 - 85
  • [25] Deepfake video detection using deep learning algorithms
    Korkmaz, Sahin
    Alkan, Mustafa
    JOURNAL OF POLYTECHNIC-POLITEKNIK DERGISI, 2023, 26 (02): : 855 - 862
  • [26] Survey of Deep Learning Algorithms for Agricultural Pest Detection
    Jiang, Xinlu
    Chen, Tianen
    Wang, Cong
    Li, Shuqin
    Zhang, Hongming
    Zhao, Chunjiang
    Computer Engineering and Applications, 2023, 59 (06) : 30 - 44
  • [27] AdvDoor: Adversarial Backdoor Attack of Deep Learning System
    Zhang, Quan
    Ding, Yifeng
    Tian, Yongqiang
    Guo, Jianmin
    Yuan, Min
    Jiang, Yu
    ISSTA '21: PROCEEDINGS OF THE 30TH ACM SIGSOFT INTERNATIONAL SYMPOSIUM ON SOFTWARE TESTING AND ANALYSIS, 2021, : 127 - 138
  • [28] TrojDRL: Evaluation of Backdoor Attacks on Deep Reinforcement Learning
    Kiourti, Panagiota
    Wardega, Kacper
    Jha, Susmit
    Li, Wenchao
    PROCEEDINGS OF THE 2020 57TH ACM/EDAC/IEEE DESIGN AUTOMATION CONFERENCE (DAC), 2020,
  • [29] Backdoor Attacks to Deep Learning Models and Countermeasures: A Survey
    Li, Yudong
    Zhang, Shigeng
    Wang, Weiping
    Song, Hong
    IEEE OPEN JOURNAL OF THE COMPUTER SOCIETY, 2023, 4 : 134 - 146
  • [30] Survey of Backdoor Attack and Defense Algorithms Based on Federated Learning
    Liu, Jialang
    Guo, Yanming
    Lao, Mingrui
    Yu, Tianyuan
    Wu, Yulun
    Feng, Yunhao
    Wu, Jiazhuang
    Jisuanji Yanjiu yu Fazhan/Computer Research and Development, 2024, 61 (10): : 2607 - 2626