Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866915006746984448 |
|---|---|
| author | Xu, Zhao Liu, Fan Liu, Hao |
| author_facet | Xu, Zhao Liu, Fan Liu, Hao |
| contents | Although Large Language Models (LLMs) have demonstrated significant capabilities in executing complex tasks in a zero-shot manner, they are susceptible to jailbreak attacks and can be manipulated to produce harmful outputs. Recently, a growing body of research has categorized jailbreak attacks into token-level and prompt-level attacks. However, previous work primarily overlooks the diverse key factors of jailbreak attacks, with most studies concentrating on LLM vulnerabilities and lacking exploration of defense-enhanced LLMs. To address these issues, we introduced $\textbf{JailTrickBench}$ to evaluate the impact of various attack settings on LLM performance and provide a baseline for jailbreak attacks, encouraging the adoption of a standardized evaluation framework. Specifically, we evaluate the eight key factors of implementing jailbreak attacks on LLMs from both target-level and attack-level perspectives. We further conduct seven representative jailbreak attacks on six defense methods across two widely used datasets, encompassing approximately 354 experiments with about 55,000 GPU hours on A800-80G. Our experimental results highlight the need for standardized benchmarking to evaluate these attacks on defense-enhanced LLMs. Our code is available at https://github.com/usail-hkust/JailTrickBench. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_09324 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs Xu, Zhao Liu, Fan Liu, Hao Cryptography and Security Artificial Intelligence Computation and Language Although Large Language Models (LLMs) have demonstrated significant capabilities in executing complex tasks in a zero-shot manner, they are susceptible to jailbreak attacks and can be manipulated to produce harmful outputs. Recently, a growing body of research has categorized jailbreak attacks into token-level and prompt-level attacks. However, previous work primarily overlooks the diverse key factors of jailbreak attacks, with most studies concentrating on LLM vulnerabilities and lacking exploration of defense-enhanced LLMs. To address these issues, we introduced $\textbf{JailTrickBench}$ to evaluate the impact of various attack settings on LLM performance and provide a baseline for jailbreak attacks, encouraging the adoption of a standardized evaluation framework. Specifically, we evaluate the eight key factors of implementing jailbreak attacks on LLMs from both target-level and attack-level perspectives. We further conduct seven representative jailbreak attacks on six defense methods across two widely used datasets, encompassing approximately 354 experiments with about 55,000 GPU hours on A800-80G. Our experimental results highlight the need for standardized benchmarking to evaluate these attacks on defense-enhanced LLMs. Our code is available at https://github.com/usail-hkust/JailTrickBench. |
| title | Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs |
| topic | Cryptography and Security Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2406.09324 |