Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2411.18000 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915038936170496 |
|---|---|
| author | Hao, Shuyang Hooi, Bryan Liu, Jun Chang, Kai-Wei Huang, Zi Cai, Yujun |
| author_facet | Hao, Shuyang Hooi, Bryan Liu, Jun Chang, Kai-Wei Huang, Zi Cai, Yujun |
| contents | Despite inheriting security measures from underlying language models, Vision-Language Models (VLMs) may still be vulnerable to safety alignment issues. Through empirical analysis, we uncover two critical findings: scenario-matched images can significantly amplify harmful outputs, and contrary to common assumptions in gradient-based attacks, minimal loss values do not guarantee optimal attack effectiveness. Building on these insights, we introduce MLAI (Multi-Loss Adversarial Images), a novel jailbreak framework that leverages scenario-aware image generation for semantic alignment, exploits flat minima theory for robust adversarial image selection, and employs multi-image collaborative attacks for enhanced effectiveness. Extensive experiments demonstrate MLAI's significant impact, achieving attack success rates of 77.75% on MiniGPT-4 and 82.80% on LLaVA-2, substantially outperforming existing methods by margins of 34.37% and 12.77% respectively. Furthermore, MLAI shows considerable transferability to commercial black-box VLMs, achieving up to 60.11% success rate. Our work reveals fundamental visual vulnerabilities in current VLMs safety mechanisms and underscores the need for stronger defenses. Warning: This paper contains potentially harmful example text. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2411_18000 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Exploring Visual Vulnerabilities via Multi-Loss Adversarial Search for Jailbreaking Vision-Language Models Hao, Shuyang Hooi, Bryan Liu, Jun Chang, Kai-Wei Huang, Zi Cai, Yujun Computer Vision and Pattern Recognition Despite inheriting security measures from underlying language models, Vision-Language Models (VLMs) may still be vulnerable to safety alignment issues. Through empirical analysis, we uncover two critical findings: scenario-matched images can significantly amplify harmful outputs, and contrary to common assumptions in gradient-based attacks, minimal loss values do not guarantee optimal attack effectiveness. Building on these insights, we introduce MLAI (Multi-Loss Adversarial Images), a novel jailbreak framework that leverages scenario-aware image generation for semantic alignment, exploits flat minima theory for robust adversarial image selection, and employs multi-image collaborative attacks for enhanced effectiveness. Extensive experiments demonstrate MLAI's significant impact, achieving attack success rates of 77.75% on MiniGPT-4 and 82.80% on LLaVA-2, substantially outperforming existing methods by margins of 34.37% and 12.77% respectively. Furthermore, MLAI shows considerable transferability to commercial black-box VLMs, achieving up to 60.11% success rate. Our work reveals fundamental visual vulnerabilities in current VLMs safety mechanisms and underscores the need for stronger defenses. Warning: This paper contains potentially harmful example text. |
| title | Exploring Visual Vulnerabilities via Multi-Loss Adversarial Search for Jailbreaking Vision-Language Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2411.18000 |