Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
[ICLR 2024] The official implementation of our ICLR2024 paper "AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models".
| Date | Stars |
|---|---|
| 2026-07-31 | 454 |
| 2026-08-01 | 454 |
| 2026-08-06 | 454 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# AutoDAN (💥 News: We have released **[AutoDAN-Turbo](https://autodans.github.io/AutoDAN-Turbo/)**, a **life-long agent** for jailbreak redteaming! It's the newest and SotA attack we have developed! Check it out! [Project](https://autodans.github.io/AutoDAN-Turbo/), [Code](https://github.com/SaFoLab-WISC/AutoDAN-Turbo), [Paper](https://arxiv.org/abs/2410.05295), [AK's recommendation](https://x.com/_akhaliq/status/1844258704633340284) ) The official implementation of our ICLR2024 paper "[AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models](https://arxiv.org/abs/2310.04451)", by *[Xiaogeng Liu](https://sheltonliu-n.github.io), [Nan Xu](https://sites.google.com/site/xunannancy/), [Muhao Chen](https://muhaochen.github.io), and [Chaowei Xiao](https://xiaocw11.github.io).*     ## Abstract The aligned Large Language Models (LLMs) are powerful language understanding and decision-making tools that are created through extensive alignment with human feedback. However, these large models remain susceptible to jailbreak attacks, where adversaries manipulate prompts to elicit malicious outputs that should not be given by aligned LLMs. Investigating jailbreak prompts can lead us to delve into the limitations of LLMs and further guide us to secure them. Unfortunately, existing jailbreak techniques suffer from either (1) scalability issues, where attacks heavily rely on manual crafting of prompts, or (2) stealthiness problems, as attacks depend on token-based algorithms to generate prompts that are often semantically meaningless, making them susceptible to detection through basic perplexity testing. In light of these challenges, we intend to answer this question: Can we develop an approach that can automatically generate stealthy jailbreak prompts? In this paper, we introduce AutoDAN, a novel jailbreak attack against aligned LLMs. AutoDAN can automatically generate stealthy jailbreak prompts by the carefully designed hierarchical genetic algorithm. Extensive evaluations demonstrate that AutoDAN not only automates the process while preserving semantic meaningfulness, but also demonstrates superior attack strength in cross-model transferability, and cross-sample universality compared with the baseline. Moreover, we also compare AutoDAN with perplexity-based defense methods and show that AutoDAN can bypass them effectively. <img src="AutoDAN.png" width="700"/> ## Latest Update | Date | Event | |------------|----------| | **2024/10/09** | 💥 We have released **[AutoDAN-Turbo](https://autodans.github.io/AutoDAN-Turbo/)**, a **life-long agent** for jailbreak redteaming! It's the newest and SotA attack we have developed! Check it out! [Project](https://autodans.github.io/AutoDAN-Turbo/), [Code](https://github.com/SaFoLab-WISC/AutoDAN-Turbo), [Paper](https://arxiv.org/abs/2410.05295) | | **2024/08/16** | 🎉 Our new [paper](https://www.usenix.org/conference/usenixsecurity24/presentation/yu-zhiyuan) on jailbreak attacks wins the **USENIX Security Distinguished Paper Award**! Don’t miss it! | | **2024/02/07** | 🔥 AutoDAN is evaluated by the [Harmbench](https://www.harmbench.org) benchmark as one of the strongest attacks. Check it out! | | **2024/02/07** | 🔥 AutoDAN is evaluated by the [Easyjailbreak](http://easyjailbreak.org) benchmark as one of the strongest attacks. Check it out! | | **2024/02/03** | We have released the full implementation of AutoDAN. Thanks to all the collaborators. | | **2024/01/16** | AutoDAN is accepted by ICLR 2024! | | **2023/10/11** | We have released a quick implementation of AutoDAN. | | **2023/10/03** | We have released
Excerpt of 5,233 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:ac8533de83f7cf8b, desc:jailbreak