FMVP: Masked Flow Matching for Adversarial Video Purification
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866909986329722880 |
|---|---|
| author | Tang, Duoxun Zhang, Xueyi Wang, Chak Hin Xiao, Xi Dai, Dasen Jiang, Xinhang Shi, Wentao Li, Rui Li, Qing |
| author_facet | Tang, Duoxun Zhang, Xueyi Wang, Chak Hin Xiao, Xi Dai, Dasen Jiang, Xinhang Shi, Wentao Li, Rui Li, Qing |
| contents | Video recognition models remain vulnerable to adversarial attacks, while existing diffusion-based purification methods suffer from inefficient sampling and curved trajectories. Directly regressing clean videos from adversarial inputs often fails to recover faithful content due to the subtle nature of perturbations; this necessitates physically shattering the adversarial structure. Therefore, we propose Flow Matching for Adversarial Video Purification FMVP. FMVP physically shatters global adversarial structures via a masking strategy and reconstructs clean video dynamics using Conditional Flow Matching (CFM) with an inpainting objective. To further decouple semantic content from adversarial noise, we design a Frequency-Gated Loss (FGL) that explicitly suppresses high-frequency adversarial residuals while preserving low-frequency fidelity. We design Attack-Aware and Generalist training paradigms to handle known and unknown threats, respectively. Extensive experiments on UCF-101 and HMDB-51 demonstrate that FMVP outperforms state-of-the-art methods (DiffPure, Defense Patterns (DP), Temporal Shuffling (TS) and FlowPure), achieving robust accuracy exceeding 87% against PGD and 89% against CW attacks. Furthermore, FMVP demonstrates superior robustness against adaptive attacks (DiffHammer) and functions as a zero-shot adversarial detector, attaining AUC-ROC scores of 0.98 for PGD and 0.79 for highly imperceptible CW attacks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_02228 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | FMVP: Masked Flow Matching for Adversarial Video Purification Tang, Duoxun Zhang, Xueyi Wang, Chak Hin Xiao, Xi Dai, Dasen Jiang, Xinhang Shi, Wentao Li, Rui Li, Qing Computer Vision and Pattern Recognition Video recognition models remain vulnerable to adversarial attacks, while existing diffusion-based purification methods suffer from inefficient sampling and curved trajectories. Directly regressing clean videos from adversarial inputs often fails to recover faithful content due to the subtle nature of perturbations; this necessitates physically shattering the adversarial structure. Therefore, we propose Flow Matching for Adversarial Video Purification FMVP. FMVP physically shatters global adversarial structures via a masking strategy and reconstructs clean video dynamics using Conditional Flow Matching (CFM) with an inpainting objective. To further decouple semantic content from adversarial noise, we design a Frequency-Gated Loss (FGL) that explicitly suppresses high-frequency adversarial residuals while preserving low-frequency fidelity. We design Attack-Aware and Generalist training paradigms to handle known and unknown threats, respectively. Extensive experiments on UCF-101 and HMDB-51 demonstrate that FMVP outperforms state-of-the-art methods (DiffPure, Defense Patterns (DP), Temporal Shuffling (TS) and FlowPure), achieving robust accuracy exceeding 87% against PGD and 89% against CW attacks. Furthermore, FMVP demonstrates superior robustness against adaptive attacks (DiffHammer) and functions as a zero-shot adversarial detector, attaining AUC-ROC scores of 0.98 for PGD and 0.79 for highly imperceptible CW attacks. |
| title | FMVP: Masked Flow Matching for Adversarial Video Purification |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.02228 |