Adversarial scratches: Deployable attacks to CNN classifiers

doi:10.1016/j.patcog.2022.108985

Adversarial scratches: Deployable attacks to CNN classifiers

2023-01-01

13

OA

AI

DOI：10.1016/j.patcog.2022.108985

Original

Save

View PDF

Abstract

En 中文

A growing body of work has shown that deep neural networks are susceptible to adversarial examples. These take the form of small perturbations applied to the model's input which lead to incorrect predictions. Unfortunately, most literature focuses on visually imperceivable perturbations to be applied to digital images that often are, by design, impossible to be deployed to physical targets. We present Adversarial Scratches: a novel L-0 black-box attack, which takes the form of scratches in images, and which possesses much greater deployability than other state-of-the-art attacks. Adversarial Scratches leverage Bezier Curves to reduce the dimension of the search space and possibly constrain the attack to a specific location. We test Adversarial Scratches in several scenarios, including a publicly available API and images of traffic signs. Results show that our attack achieves higher fooling rate than other deployable state-of-the-art methods, while requiring significantly fewer queries and modifying very few pixels. (C) 2022 Elsevier Ltd. All rights reserved.

Keywords:

Adversarial perturbations

Adversarial attacks

Deep learning

Convolutional neural networks

Bezier curves

Adversarial scratches: Deployable attacks to CNN classifiers

Abstract

AI Summary

Journal

Organization