Position: Towards Resilience Against Adversarial Examples

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Dai, Sihui, Xiang, Chong, Wu, Tong, Mittal, Prateek
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912063594430464
author Dai, Sihui
Xiang, Chong
Wu, Tong
Mittal, Prateek
author_facet Dai, Sihui
Xiang, Chong
Wu, Tong
Mittal, Prateek
contents Current research on defending against adversarial examples focuses primarily on achieving robustness against a single attack type such as $\ell_2$ or $\ell_{\infty}$-bounded attacks. However, the space of possible perturbations is much larger than considered by many existing defenses and is difficult to mathematically model, so the attacker can easily bypass the defense by using a type of attack that is not covered by the defense. In this position paper, we argue that in addition to robustness, we should also aim to develop defense algorithms that are adversarially resilient -- defense algorithms should specify a means to quickly adapt the defended model to be robust against new attacks. We provide a definition of adversarial resilience and outline considerations of designing an adversarially resilient defense. We then introduce a subproblem of adversarial resilience which we call continual adaptive robustness, in which the defender gains knowledge of the formulation of possible perturbation spaces over time and can then update their model based on this information. Additionally, we demonstrate the connection between continual adaptive robustness and previously studied problems of multiattack robustness and unforeseen attack robustness and outline open directions within these fields which can contribute to improving continual adaptive robustness and adversarial resilience.
format Preprint
id arxiv_https___arxiv_org_abs_2405_01349
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Position: Towards Resilience Against Adversarial Examples
Dai, Sihui
Xiang, Chong
Wu, Tong
Mittal, Prateek
Machine Learning
Cryptography and Security
Current research on defending against adversarial examples focuses primarily on achieving robustness against a single attack type such as $\ell_2$ or $\ell_{\infty}$-bounded attacks. However, the space of possible perturbations is much larger than considered by many existing defenses and is difficult to mathematically model, so the attacker can easily bypass the defense by using a type of attack that is not covered by the defense. In this position paper, we argue that in addition to robustness, we should also aim to develop defense algorithms that are adversarially resilient -- defense algorithms should specify a means to quickly adapt the defended model to be robust against new attacks. We provide a definition of adversarial resilience and outline considerations of designing an adversarially resilient defense. We then introduce a subproblem of adversarial resilience which we call continual adaptive robustness, in which the defender gains knowledge of the formulation of possible perturbation spaces over time and can then update their model based on this information. Additionally, we demonstrate the connection between continual adaptive robustness and previously studied problems of multiattack robustness and unforeseen attack robustness and outline open directions within these fields which can contribute to improving continual adaptive robustness and adversarial resilience.
title Position: Towards Resilience Against Adversarial Examples
topic Machine Learning
Cryptography and Security
url https://arxiv.org/abs/2405.01349