Saved in:
Bibliographic Details
Main Authors: Su, Linxiang, Balogh, András
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.17362
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911568157999104
author Su, Linxiang
Balogh, András
author_facet Su, Linxiang
Balogh, András
contents Despite its remarkable success in zero-shot image-text matching, CLIP remains highly vulnerable to adversarial perturbations on images. As adversarial fine-tuning is prohibitively costly, recent works explore various test-time defense strategies; however, these approaches still exhibit limited robustness. In this work, we revisit this problem and propose a simple yet effective strategy: Augmentation-based Test-time Adversarial Correction (ATAC). Our method operates directly in the embedding space of CLIP, calculating augmentation-induced drift vectors to infer a semantic recovery direction and correcting the embedding based on the angular consistency of these latent drifts. Across a wide range of benchmarks, ATAC consistently achieves remarkably high robustness, surpassing that of previous state-of-the-art methods by nearly 50\% on average, all while requiring minimal computational overhead. Furthermore, ATAC retains state-of-the-art robustness in unconventional and extreme settings and even achieves nontrivial robustness against adaptive attacks. Our results demonstrate that ATAC is an efficient method in a novel paradigm for test-time adversarial defenses in the embedding space of CLIP. Code is available at: https://github.com/kylin0421/ATAC
format Preprint
id arxiv_https___arxiv_org_abs_2511_17362
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ATAC: Augmentation-Based Test-Time Adversarial Correction for CLIP
Su, Linxiang
Balogh, András
Computer Vision and Pattern Recognition
Despite its remarkable success in zero-shot image-text matching, CLIP remains highly vulnerable to adversarial perturbations on images. As adversarial fine-tuning is prohibitively costly, recent works explore various test-time defense strategies; however, these approaches still exhibit limited robustness. In this work, we revisit this problem and propose a simple yet effective strategy: Augmentation-based Test-time Adversarial Correction (ATAC). Our method operates directly in the embedding space of CLIP, calculating augmentation-induced drift vectors to infer a semantic recovery direction and correcting the embedding based on the angular consistency of these latent drifts. Across a wide range of benchmarks, ATAC consistently achieves remarkably high robustness, surpassing that of previous state-of-the-art methods by nearly 50\% on average, all while requiring minimal computational overhead. Furthermore, ATAC retains state-of-the-art robustness in unconventional and extreme settings and even achieves nontrivial robustness against adaptive attacks. Our results demonstrate that ATAC is an efficient method in a novel paradigm for test-time adversarial defenses in the embedding space of CLIP. Code is available at: https://github.com/kylin0421/ATAC
title ATAC: Augmentation-Based Test-Time Adversarial Correction for CLIP
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.17362