Adversarial Prompt Tuning for Vision-Language Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2023
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866911991759634432 |
|---|---|
| author | Zhang, Jiaming Ma, Xingjun Wang, Xin Qiu, Lingyu Wang, Jiaqi Jiang, Yu-Gang Sang, Jitao |
| author_facet | Zhang, Jiaming Ma, Xingjun Wang, Xin Qiu, Lingyu Wang, Jiaqi Jiang, Yu-Gang Sang, Jitao |
| contents | With the rapid advancement of multimodal learning, pre-trained Vision-Language Models (VLMs) such as CLIP have demonstrated remarkable capacities in bridging the gap between visual and language modalities. However, these models remain vulnerable to adversarial attacks, particularly in the image modality, presenting considerable security risks. This paper introduces Adversarial Prompt Tuning (AdvPT), a novel technique to enhance the adversarial robustness of image encoders in VLMs. AdvPT innovatively leverages learnable text prompts and aligns them with adversarial image embeddings, to address the vulnerabilities inherent in VLMs without the need for extensive parameter training or modification of the model architecture. We demonstrate that AdvPT improves resistance against white-box and black-box adversarial attacks and exhibits a synergistic effect when combined with existing image-processing-based defense techniques, further boosting defensive capabilities. Comprehensive experimental analyses provide insights into adversarial prompt tuning, a novel paradigm devoted to improving resistance to adversarial images through textual input modifications, paving the way for future robust multimodal learning research. These findings open up new possibilities for enhancing the security of VLMs. Our code is available at https://github.com/jiamingzhang94/Adversarial-Prompt-Tuning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2311_11261 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Adversarial Prompt Tuning for Vision-Language Models Zhang, Jiaming Ma, Xingjun Wang, Xin Qiu, Lingyu Wang, Jiaqi Jiang, Yu-Gang Sang, Jitao Computer Vision and Pattern Recognition Artificial Intelligence With the rapid advancement of multimodal learning, pre-trained Vision-Language Models (VLMs) such as CLIP have demonstrated remarkable capacities in bridging the gap between visual and language modalities. However, these models remain vulnerable to adversarial attacks, particularly in the image modality, presenting considerable security risks. This paper introduces Adversarial Prompt Tuning (AdvPT), a novel technique to enhance the adversarial robustness of image encoders in VLMs. AdvPT innovatively leverages learnable text prompts and aligns them with adversarial image embeddings, to address the vulnerabilities inherent in VLMs without the need for extensive parameter training or modification of the model architecture. We demonstrate that AdvPT improves resistance against white-box and black-box adversarial attacks and exhibits a synergistic effect when combined with existing image-processing-based defense techniques, further boosting defensive capabilities. Comprehensive experimental analyses provide insights into adversarial prompt tuning, a novel paradigm devoted to improving resistance to adversarial images through textual input modifications, paving the way for future robust multimodal learning research. These findings open up new possibilities for enhancing the security of VLMs. Our code is available at https://github.com/jiamingzhang94/Adversarial-Prompt-Tuning. |
| title | Adversarial Prompt Tuning for Vision-Language Models |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2311.11261 |