Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916270412136448 |
|---|---|
| author | Gu, Xiangming Zheng, Xiaosen Pang, Tianyu Du, Chao Liu, Qian Wang, Ye Jiang, Jing Lin, Min |
| author_facet | Gu, Xiangming Zheng, Xiaosen Pang, Tianyu Du, Chao Liu, Qian Wang, Ye Jiang, Jing Lin, Min |
| contents | A multimodal large language model (MLLM) agent can receive instructions, capture images, retrieve histories from memory, and decide which tools to use. Nonetheless, red-teaming efforts have revealed that adversarial images/prompts can jailbreak an MLLM and cause unaligned behaviors. In this work, we report an even more severe safety issue in multi-agent environments, referred to as infectious jailbreak. It entails the adversary simply jailbreaking a single agent, and without any further intervention from the adversary, (almost) all agents will become infected exponentially fast and exhibit harmful behaviors. To validate the feasibility of infectious jailbreak, we simulate multi-agent environments containing up to one million LLaVA-1.5 agents, and employ randomized pair-wise chat as a proof-of-concept instantiation for multi-agent interaction. Our results show that feeding an (infectious) adversarial image into the memory of any randomly chosen agent is sufficient to achieve infectious jailbreak. Finally, we derive a simple principle for determining whether a defense mechanism can provably restrain the spread of infectious jailbreak, but how to design a practical defense that meets this principle remains an open question to investigate. Our project page is available at https://sail-sg.github.io/Agent-Smith/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_08567 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast Gu, Xiangming Zheng, Xiaosen Pang, Tianyu Du, Chao Liu, Qian Wang, Ye Jiang, Jing Lin, Min Computation and Language Cryptography and Security Computer Vision and Pattern Recognition Machine Learning Multiagent Systems A multimodal large language model (MLLM) agent can receive instructions, capture images, retrieve histories from memory, and decide which tools to use. Nonetheless, red-teaming efforts have revealed that adversarial images/prompts can jailbreak an MLLM and cause unaligned behaviors. In this work, we report an even more severe safety issue in multi-agent environments, referred to as infectious jailbreak. It entails the adversary simply jailbreaking a single agent, and without any further intervention from the adversary, (almost) all agents will become infected exponentially fast and exhibit harmful behaviors. To validate the feasibility of infectious jailbreak, we simulate multi-agent environments containing up to one million LLaVA-1.5 agents, and employ randomized pair-wise chat as a proof-of-concept instantiation for multi-agent interaction. Our results show that feeding an (infectious) adversarial image into the memory of any randomly chosen agent is sufficient to achieve infectious jailbreak. Finally, we derive a simple principle for determining whether a defense mechanism can provably restrain the spread of infectious jailbreak, but how to design a practical defense that meets this principle remains an open question to investigate. Our project page is available at https://sail-sg.github.io/Agent-Smith/. |
| title | Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast |
| topic | Computation and Language Cryptography and Security Computer Vision and Pattern Recognition Machine Learning Multiagent Systems |
| url | https://arxiv.org/abs/2402.08567 |