Two Heads Are Better Than One: Averaging along Fine-Tuning to Improve Targeted Transferability
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913629250519040 |
|---|---|
| author | Zeng, Hui Cui, Sanshuai Chen, Biwei Peng, Anjie |
| author_facet | Zeng, Hui Cui, Sanshuai Chen, Biwei Peng, Anjie |
| contents | With much longer optimization time than that of untargeted attacks notwithstanding, the transferability of targeted attacks is still far from satisfactory. Recent studies reveal that fine-tuning an existing adversarial example (AE) in feature space can efficiently boost its targeted transferability. However, existing fine-tuning schemes only utilize the endpoint and ignore the valuable information in the fine-tuning trajectory. Noting that the vanilla fine-tuning trajectory tends to oscillate around the periphery of a flat region of the loss surface, we propose averaging over the fine-tuning trajectory to pull the crafted AE towards a more centered region. We compare the proposed method with existing fine-tuning schemes by integrating them with state-of-the-art targeted attacks in various attacking scenarios. Experimental results uphold the superiority of the proposed method in boosting targeted transferability. The code is available at github.com/zengh5/Avg_FT. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_20807 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Two Heads Are Better Than One: Averaging along Fine-Tuning to Improve Targeted Transferability Zeng, Hui Cui, Sanshuai Chen, Biwei Peng, Anjie Computer Vision and Pattern Recognition Artificial Intelligence With much longer optimization time than that of untargeted attacks notwithstanding, the transferability of targeted attacks is still far from satisfactory. Recent studies reveal that fine-tuning an existing adversarial example (AE) in feature space can efficiently boost its targeted transferability. However, existing fine-tuning schemes only utilize the endpoint and ignore the valuable information in the fine-tuning trajectory. Noting that the vanilla fine-tuning trajectory tends to oscillate around the periphery of a flat region of the loss surface, we propose averaging over the fine-tuning trajectory to pull the crafted AE towards a more centered region. We compare the proposed method with existing fine-tuning schemes by integrating them with state-of-the-art targeted attacks in various attacking scenarios. Experimental results uphold the superiority of the proposed method in boosting targeted transferability. The code is available at github.com/zengh5/Avg_FT. |
| title | Two Heads Are Better Than One: Averaging along Fine-Tuning to Improve Targeted Transferability |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2412.20807 |