arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2606.11202cs.CL
收藏

一次越狱,多种语言:学习语言无关的意图表示用于多语言越狱检测

One Jailbreak, Many Tongues: Learning Language-Insensitive Intention Representations for Multilingual Jailbreak Detection

School of Cyber Science and Engineering, Sichuan University(四川大学网络空间安全学院);School of Computer Science and Engineering, Nanyang Technological University(南洋理工大学计算机科学与工程学院);School of Computer Science and Engineering, University of Electronic Science and Technology of China(电子科技大学计算机科学与工程学院)

Shuyu Jiang, Kaiyu Xu, Xingshu Chen, Hao Ren, Rui Tang, Yi Zhang, Tianwei Zhang, Hongwei Li

AI总结:

针对多语言LLM安全漏洞,提出MLJailDe框架,通过多语言回译数据增强和相对距离约束,实现跨语言越狱检测,F1达98.5%。

摘要:

大型语言模型(LLMs)越来越多地部署在面向全球多语言用户的应用程序中,然而安全训练仍集中在主流语言上,并未与多语言能力同步发展,从而为越狱攻击创造了可利用的漏洞。当前的越狱防御主要是在主流语言中开发和评估的,其有效性受到对齐的多语言监督稀缺以及语言变异导致的表示分散的限制。为了解决这个问题,我们提出了MLJailDe,一个多语言越狱检测框架,旨在提高多语言鲁棒性和跨语言泛化能力。MLJailDe首先引入了一种多语言回译数据增强算法,构建了一个语义一致且功能有效的数据集,涵盖11种语言,包含2,232个良性样本和1,239个越狱样本。在此基础上,MLJailDe采用相对距离约束来减少跨语言表示分散,并鼓励具有相似意图的越狱提示在不同语言中形成一致的聚类,同时进一步使用不平衡感知的分类目标来缓解类别不平衡并学习更可靠的多语言决策边界。实验结果表明,MLJailDe在多种语言上优于最先进的基线,F1分数达到98.5%,并且在未见过的语言上平均F1分数达到97.1%,展示了强大的有效性和跨语言泛化能力。

英文摘要:

Large language models (LLMs) are increasingly deployed in applications for global multilingual users, yet safety training remains concentrated in dominant languages and has not progressed in parallel with multilingual capability, creating exploitable gaps for jailbreak attacks. Current jailbreak defenses are largely developed and evaluated in dominant languages, and their effectiveness is limited by the scarcity of aligned multilingual supervision and representations dispersion caused by language variation. To address this issue, we propose MLJailDe, a multilingual jailbreak detection framework designed to improve both multilingual robustness and cross-lingual generalization. MLJailDe first introduces a multilingual back-translation data augmentation algorithm to construct a semantically consistent and functionally effective dataset spanning 11 languages, consisting of 2,232 benign and 1,239 jailbreak samples. On this basis, MLJailDe employs relative-distance constraints to reduce cross-lingual representation dispersion and encourage jailbreak prompts with similar intent to form consistent clusters across languages, while an imbalance-aware classification objective is further used to alleviate class imbalance and learn more reliable multilingual decision boundaries. Experimental results show that MLJailDe outperforms state-of-the-art baselines across multiple languages, achieving an F1 score of 98.5\%, and obtains an average F1 score of 97.1\% on unseen languages, demonstrating strong effectiveness and cross-lingual generalization.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑