Stochastic sparse adversarial attacks

Manon Césaire, Lucas Schott, Hatem Hajri, Sylvain Lamprier, Patrick Gallinari
11/24/2020
cs.LGcs.CRcs.CV

Official code: SSAA3/stochastic-sparse-adv-attacks (3 stars, last commit 2022-06-27).

Abstract

This paper introduces stochastic sparse adversarial attacks (SSAA), standing as simple, fast and purely noise-based targeted and untargeted attacks of neural network classifiers (NNC). SSAA offer new examples of sparse (or $L_0$) attacks for which only few methods have been proposed previously. These attacks are devised by exploiting a small-time expansion idea widely used for Markov processes. Experiments on small and large datasets (CIFAR-10 and ImageNet) illustrate several advantages of SSAA in comparison with the-state-of-the-art methods. For instance, in the untargeted case, our method called Voting Folded Gaussian Attack (VFGA) scales efficiently to ImageNet and achieves a significantly lower $L_0$ score than SparseFool (up to $\frac{2}{5}$) while being faster. Moreover, VFGA achieves better $L_0$ scores on ImageNet than Sparse-RS when both attacks are fully successful on a large number of samples.

View on arXivView PDF

Code Implementations(1)

SSAA3/stochastic-sparse-adv-attacksOfficial100%

Stochastic Sparse Adversarial Attacks

33PythonJul 19, 20214 years ago
adversarial-attacksrandom-noisessparse-attacks

Similar Papers

Combating Adversarial Attacks Using Sparse Representations

87% match1 repo72018

Sparsity-based Defense against Adversarial Attacks on Linear Classifiers

86% match1 repo72018

Towards Interpretable Adversarial Examples via Sparse Adversarial Attack

86% match1 repo12025

Smoothed Inference for Adversarially-Trained Models

85% match1 repo32019

A Versatile Framework for Designing Group-Sparse Adversarial Attacks

85% match2025

Sparse but Strong: Crafting Adversarially Robust Graph Lottery Tickets

84% match2023

Cite this paper

@article{csaire2020stochastic,
  title  = {Stochastic sparse adversarial attacks},
  author = {Manon Césaire and Lucas Schott and Hatem Hajri and Sylvain Lamprier and Patrick Gallinari},
  year   = {2020},
  eprint = {2011.12423},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2011.12423}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2011.12423.svg)](https://codeofpaper.com/paper/2011.12423)

Discussion