Corpus ID: 67788180

Wasserstein Adversarial Examples via Projected Sinkhorn Iterations

@article{Wong2019WassersteinAE,
  title={Wasserstein Adversarial Examples via Projected Sinkhorn Iterations},
  author={Eric Wong and Frank R. Schmidt and J. Zico Kolter},
  journal={ArXiv},
  year={2019},
  volume={abs/1902.07906}
}
  • Eric Wong, Frank R. Schmidt, J. Zico Kolter
  • Published 2019
  • Mathematics, Computer Science
  • ArXiv
  • A rapidly growing area of work has studied the existence of adversarial examples, datapoints which have been perturbed to fool a classifier, but the vast majority of these works have focused primarily on threat models defined by $\ell_p$ norm-bounded perturbations. In this paper, we propose a new threat model for adversarial attacks based on the Wasserstein distance. In the image classification setting, such distances measure the cost of moving pixel mass, which naturally cover "standard" image… CONTINUE READING

    Citations

    Publications citing this paper.
    SHOWING 1-10 OF 45 CITATIONS

    Wasserstein Smoothing: Certified Robustness against Wasserstein Adversarial Attacks

    VIEW 26 EXCERPTS
    CITES BACKGROUND, METHODS & RESULTS
    HIGHLY INFLUENCED

    Generating Adversarial Samples With Constrained Wasserstein Distance

    VIEW 4 EXCERPTS
    CITES METHODS & BACKGROUND
    HIGHLY INFLUENCED

    Functional Adversarial Attacks

    VIEW 2 EXCERPTS
    CITES METHODS & BACKGROUND

    SemanticAdv: Generating Adversarial Examples via Attribute-conditional Image Editing

    VIEW 1 EXCERPT
    CITES BACKGROUND

    Improved Image Wasserstein Attacks and Defenses

    VIEW 31 EXCERPTS
    CITES BACKGROUND, METHODS & RESULTS
    HIGHLY INFLUENCED

    Adversarial Examples on Object Recognition

    VIEW 2 EXCERPTS
    CITES METHODS & BACKGROUND

    Overfitting in adversarially robust deep learning

    VIEW 2 EXCERPTS
    CITES BACKGROUND

    Luring of Adversarial Perturbations

    VIEW 1 EXCERPT
    CITES BACKGROUND

    FILTER CITATIONS BY YEAR

    2018
    2020

    CITATION STATISTICS

    • 4 Highly Influenced Citations

    • Averaged 15 Citations per year from 2018 through 2020

    • 59% Increase in citations per year in 2020 over 2019

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 34 REFERENCES

    Spatially Transformed Adversarial Examples

    VIEW 1 EXCERPT

    Robust Physical-World Attacks on Deep Learning Visual Classification

    VIEW 1 EXCERPT

    Adversarial examples in the physical world

    VIEW 2 EXCERPTS