OMG-Attack: Self-Supervised On-Manifold Generation of Transferable Evasion Attacks

Ofir Bar Tal*, Adi Haviv, Amit H. Bermano

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Evasion Attacks (EA) are used to test the robustness of trained neural networks by distorting input data to misguide the model into incorrect classifications. Creating these attacks is a challenging task, especially with the ever increasing complexity of models and datasets. In this work, we introduce a self-supervised, computationally economical method for generating adversarial examples, designed for the unseen black-box setting. Adapting techniques from representation learning, our method generates on-manifold EAs that are encouraged to resemble the data distribution. These attacks are comparable in effectiveness compared to the state-of-the-art when attacking the model trained on, but are significantly more effective when attacking unseen models, as the attacks are more related to the data rather than the model itself. Our experiments consistently demonstrate the method is effective across various models, unseen data categories, and even defended models, suggesting a significant role for on-manifold EAs when targeting unseen models.

Original languageEnglish
Title of host publicationProceedings - 2023 IEEE/CVF International Conference on Computer Vision Workshops, ICCVW 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages3698-3708
Number of pages11
ISBN (Electronic)9798350307443
DOIs
StatePublished - 2023
Event2023 IEEE/CVF International Conference on Computer Vision Workshops, ICCVW 2023 - Paris, France
Duration: 2 Oct 20236 Oct 2023

Publication series

NameProceedings - 2023 IEEE/CVF International Conference on Computer Vision Workshops, ICCVW 2023

Conference

Conference2023 IEEE/CVF International Conference on Computer Vision Workshops, ICCVW 2023
Country/TerritoryFrance
CityParis
Period2/10/236/10/23

Keywords

  • Adversarial Attacks
  • Computer Vision
  • Contrastive Learning
  • Evasion Attacks
  • Generative Attacks
  • Generative Model
  • On Manifold
  • Robustness
  • Self Supervised
  • Transferable Attacks

Fingerprint

Dive into the research topics of 'OMG-Attack: Self-Supervised On-Manifold Generation of Transferable Evasion Attacks'. Together they form a unique fingerprint.

Cite this