MasterKey: Automated Jailbreak Across Multiple Large Language Model Chatbots

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Deng, Gelei, Liu, Yi, Li, Yuekang, Wang, Kailong, Zhang, Ying, Li, Zefeng, Wang, Haoyu, Zhang, Tianwei, Liu, Yang
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916122883784704
author Deng, Gelei
Liu, Yi
Li, Yuekang
Wang, Kailong
Zhang, Ying
Li, Zefeng
Wang, Haoyu
Zhang, Tianwei
Liu, Yang
author_facet Deng, Gelei
Liu, Yi
Li, Yuekang
Wang, Kailong
Zhang, Ying
Li, Zefeng
Wang, Haoyu
Zhang, Tianwei
Liu, Yang
contents Large Language Models (LLMs) have revolutionized Artificial Intelligence (AI) services due to their exceptional proficiency in understanding and generating human-like text. LLM chatbots, in particular, have seen widespread adoption, transforming human-machine interactions. However, these LLM chatbots are susceptible to "jailbreak" attacks, where malicious users manipulate prompts to elicit inappropriate or sensitive responses, contravening service policies. Despite existing attempts to mitigate such threats, our research reveals a substantial gap in our understanding of these vulnerabilities, largely due to the undisclosed defensive measures implemented by LLM service providers. In this paper, we present Jailbreaker, a comprehensive framework that offers an in-depth understanding of jailbreak attacks and countermeasures. Our work makes a dual contribution. First, we propose an innovative methodology inspired by time-based SQL injection techniques to reverse-engineer the defensive strategies of prominent LLM chatbots, such as ChatGPT, Bard, and Bing Chat. This time-sensitive approach uncovers intricate details about these services' defenses, facilitating a proof-of-concept attack that successfully bypasses their mechanisms. Second, we introduce an automatic generation method for jailbreak prompts. Leveraging a fine-tuned LLM, we validate the potential of automated jailbreak generation across various commercial LLM chatbots. Our method achieves a promising average success rate of 21.58%, significantly outperforming the effectiveness of existing techniques. We have responsibly disclosed our findings to the concerned service providers, underscoring the urgent need for more robust defenses. Jailbreaker thus marks a significant step towards understanding and mitigating jailbreak threats in the realm of LLM chatbots.
format Preprint
id arxiv_https___arxiv_org_abs_2307_08715
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle MasterKey: Automated Jailbreak Across Multiple Large Language Model Chatbots
Deng, Gelei
Liu, Yi
Li, Yuekang
Wang, Kailong
Zhang, Ying
Li, Zefeng
Wang, Haoyu
Zhang, Tianwei
Liu, Yang
Cryptography and Security
Large Language Models (LLMs) have revolutionized Artificial Intelligence (AI) services due to their exceptional proficiency in understanding and generating human-like text. LLM chatbots, in particular, have seen widespread adoption, transforming human-machine interactions. However, these LLM chatbots are susceptible to "jailbreak" attacks, where malicious users manipulate prompts to elicit inappropriate or sensitive responses, contravening service policies. Despite existing attempts to mitigate such threats, our research reveals a substantial gap in our understanding of these vulnerabilities, largely due to the undisclosed defensive measures implemented by LLM service providers. In this paper, we present Jailbreaker, a comprehensive framework that offers an in-depth understanding of jailbreak attacks and countermeasures. Our work makes a dual contribution. First, we propose an innovative methodology inspired by time-based SQL injection techniques to reverse-engineer the defensive strategies of prominent LLM chatbots, such as ChatGPT, Bard, and Bing Chat. This time-sensitive approach uncovers intricate details about these services' defenses, facilitating a proof-of-concept attack that successfully bypasses their mechanisms. Second, we introduce an automatic generation method for jailbreak prompts. Leveraging a fine-tuned LLM, we validate the potential of automated jailbreak generation across various commercial LLM chatbots. Our method achieves a promising average success rate of 21.58%, significantly outperforming the effectiveness of existing techniques. We have responsibly disclosed our findings to the concerned service providers, underscoring the urgent need for more robust defenses. Jailbreaker thus marks a significant step towards understanding and mitigating jailbreak threats in the realm of LLM chatbots.
title MasterKey: Automated Jailbreak Across Multiple Large Language Model Chatbots
topic Cryptography and Security
url https://arxiv.org/abs/2307.08715