arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.16686cs.CL
收藏

大型语言模型能否自动越狱GPT-4V?

Can Large Language Models Automatically Jailbreak GPT-4V?

Huazhong University of Science and Technology(华中科技大学);Lehigh University(里海大学);University of Notre Dame(圣母大学)

Yuanwei Wu, Yue Huang, Yixin Liu, Xiang Li, Pan Zhou, Lichao Sun

AI总结:

研究提出AutoJailbreak技术,利用大型语言模型进行红队测试和由弱到强的上下文学习优化越狱提示,结合早期停止搜索方法,实现对GPT-4V超过95.3%的攻击成功率,揭示了其安全漏洞。

摘要:

GPT-4V因其整合和处理多模态信息的卓越能力而备受关注。同时,其人脸识别能力引发了关于隐私泄露的新的安全担忧。尽管研究人员通过RLHF或预处理过滤器努力进行安全对齐,漏洞仍可能被利用。在我们的研究中,我们引入了AutoJailbreak,这是一种受提示优化启发的创新自动越狱技术。我们利用大型语言模型(LLMs)进行红队测试以优化越狱提示,并采用由弱到强的上下文学习提示来提升效率。此外,我们提出了一种结合早期停止的有效搜索方法,以最小化优化时间和token消耗。我们的实验表明,AutoJailbreak显著超越传统方法,实现了超过95.3%的攻击成功率(ASR)。本研究为加强GPT-4V安全性提供了启示,强调了LLMs在破坏GPT-4V完整性方面被利用的潜力。

英文摘要:

GPT-4V has attracted considerable attention due to its extraordinary capacity for integrating and processing multimodal information. At the same time, its ability of face recognition raises new safety concerns of privacy leakage. Despite researchers' efforts in safety alignment through RLHF or preprocessing filters, vulnerabilities might still be exploited. In our study, we introduce AutoJailbreak, an innovative automatic jailbreak technique inspired by prompt optimization. We leverage Large Language Models (LLMs) for red-teaming to refine the jailbreak prompt and employ weak-to-strong in-context learning prompts to boost efficiency. Furthermore, we present an effective search method that incorporates early stopping to minimize optimization time and token expenditure. Our experiments demonstrate that AutoJailbreak significantly surpasses conventional methods, achieving an Attack Success Rate (ASR) exceeding 95.3\%. This research sheds light on strengthening GPT-4V security, underscoring the potential for LLMs to be exploited in compromising GPT-4V integrity.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑