Leveraging Attack Non-Transferability to Boost Adversarial Robustness for Foundation Models
Applied Sciences · 17.04.2026 · 16 (8) · 3894
DOI: 10.3390/app16083894 ↗Kurz gefasst
Fachartikel in der Zeitschrift Applied Sciences, online erschienen am 17.04.2026. Verfasst von Koshiro Toishi und 4 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: JSPS KAKENHI.
Abstract (Originalsprache)
This paper presents a novel adversarial defense framework that strategically exploits the non-transferability of adversarial attacks across multi-modal foundation models. While Contrastive Language–Image Pre-training (CLIP) models demonstrate remarkable zero-shot capabilities, they remain vulnerable to adversarial samples. Adversarial fine-tuning is widely adopted as a standard defense, yet the resulting robustness against sophisticated white-box attacks is often insufficient. To address this limitation, we aim to boost the robustness of an adversarially fine-tuned model by utilizing a pre-trained auxiliary model to leverage attack non-transferability. Specifically, we construct a common embedding space and introduce a detection scheme that identifies the attack target based on feature distances. By adaptively switching the prediction output, we effectively mitigate attacks. Experimental results demonstrate that our approach outperforms state-of-the-art adversarial fine-tuning methods in terms of adversarial robustness.
Förderung
- JSPS KAKENHI
Zitieren
APA
Toishi, K., Maeda, K., Togo, R., Ogawa, T., & Haseyama, M. (2026). Leveraging Attack Non-Transferability to Boost Adversarial Robustness for Foundation Models. Applied Sciences, 16(8), 3894. https://doi.org/10.3390/app16083894
Vancouver
Toishi K, Maeda K, Togo R, Ogawa T, Haseyama M. Leveraging Attack Non-Transferability to Boost Adversarial Robustness for Foundation Models. Applied Sciences. 2026;16(8):3894. doi:10.3390/app16083894
BibTeX
@article{Toishi2026,
author = {Toishi, Koshiro and Maeda, Keisuke and Togo, Ren and Ogawa, Takahiro and Haseyama, Miki},
title = {Leveraging Attack Non-Transferability to Boost Adversarial Robustness for Foundation Models},
journal = {Applied Sciences},
year = {2026},
volume = {16},
number = {8},
pages = {3894},
doi = {10.3390/app16083894}
}
EUVIMED