The integration of additional modalities increases the susceptibility of large vision-language models (LVLMs) to safety risks, such as jailbreak attacks, compared to their language-only counterparts. While existing research primarily focuses on post-hoc alignment techniques, the underlying safety mechanisms within LVLMs remain largely unexplored. In this work , we investigate whether LVLMs inherently encode safety-relevant signals within their internal activations during inference. Our findings reveal that LVLMs exhibit distinct activation patterns when processing unsafe prompts, which can be leveraged to detect and mitigate adversarial inputs without requiring extensive fine-tuning. Building on this insight, we introduce HiddenDetect, a novel tuning-free framework that harnesses internal model activations to enhance safety. Experimental results show that {HiddenDetect} surpasses state-of-the-art methods in detecting jailbreak attacks against LVLMs. By utilizing intrinsic safety-aware patterns, our method provides an efficient and scalable solution for strengthening LVLM robustness against multimodal threats. Our code will be released publicly at https://github.com/leigest519/HiddenDetect.
Cross-modality Information Check for Detecting Jailbreaking in Multimodal Large Language Models
School of Information Science and Technology, ShanghaiTech University(上海科技大学信息科学与技术学院);The State Key Laboratory of Blockchain and Data Security, Zhejiang University(浙江大学区块链与数据安全国家重点实验室)
What Intermediate Layers Know: Detecting Jailbreaks from Entropy Dynamics
LMU Munich(慕尼黑大学);relAI – Konrad Zuse School of Excellence in Reliable AI(relAI – 康拉德·楚泽可靠人工智能卓越学校);University of Pisa(比萨大学);Munich Center for Machine Learning(慕尼黑机器学习中心);School of Engineering, Computing and Mathematics, Oxford Brookes University(牛津布鲁克斯大学工程、计算与数学学院)
Divide and Conquer: A Hybrid Strategy Defeats Multimodal Large Language Models
School of Software, Henan University(河南大学软件学院);Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);University of Chinese Academy of Sciences(中国科学院大学)
School of Information Management, Wuhan University(武汉大学信息管理学院);Nanyang Technological University(南洋理工大学);Torr Vision Group, University of Oxford(牛津大学Torr视觉组)
SafeDialBench: A Fine-Grained Safety Evaluation Benchmark for Large Language Models in Multi-Turn Dialogues with Diverse Jailbreak Attacks
National Key Laboratory for Novel Software Technology, Nanjing University(南京大学新型软件技术国家重点实验室);University of Liverpool(利物浦大学);City University of Hong Kong(香港城市大学);School of Intelligence Science and Technology, Nanjing University(南京大学智能科学与技术学院);China Mobile Research Institute(中国移动研究院);China Mobile (Suzhou) Software Technology Co., Ltd(中国移动苏州软件技术有限公司)
RigorLLM: Resilient Guardrails for Large Language Models against Undesired Content
University of Illinois Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校);Virginia Tech(弗吉尼亚理工大学);Salesforce Research(Salesforce研究院);University of California Berkeley(加州大学伯克利分校);University of Chicago(芝加哥大学)