Optimal transport as a defense against adversarial attacks - Archive ouverte HAL Access content directly
Conference Papers Year :

Optimal transport as a defense against adversarial attacks

(1) , (2) , (2)
1
2

Abstract

Deep learning classifiers are now known to have flaws in the representations of their class. Adversarial attacks can find a human-imperceptible perturbation for a given image that will mislead a trained model. The most effective methods to defend against such attacks trains on generated adversarial examples to learn their distribution. Previous work aimed to align original and adversarial image representations in the same way as domain adaptation to improve robustness. Yet, they partially align the representations using approaches that do not reflect the geometry of space and distribution. In addition, it is difficult to accurately compare robustness between defended models. Until now, they have been evaluated using a fixed perturbation size. However, defended models may react differently to variations of this perturbation size. In this paper, the analogy of domain adaptation is taken a step further by exploiting optimal transport theory. We propose to use a loss between distributions that faithfully reflect the ground distance. This leads to SAT (Sinkhorn Adversarial Training), a more robust defense against adversarial attacks. Then, we propose to quantify more precisely the robustness of a model to adversarial attacks over a wide range of perturbation sizes using a different metric, the Area Under the Accuracy Curve (AUAC). We perform extensive experiments on both CIFAR-10 and CIFAR-100 datasets and show that our defense is globally more robust than the state-of-the-art.
Fichier principal
Vignette du fichier
Sinkhorn_Defense_v2.pdf (420.98 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

cea-03251815 , version 1 (07-06-2021)

Identifiers

  • HAL Id : cea-03251815 , version 1

Cite

Quentin Bouniot, Romaric Audigier, Angelique Loesch. Optimal transport as a defense against adversarial attacks. ICPR 2020 - 25th International Conference on Pattern Recognition, Jan 2021, Milano (Virtual conference), Italy. ⟨cea-03251815⟩
32 View
61 Download

Share

Gmail Facebook Twitter LinkedIn More