arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.06854cs.CL
收藏

SEMA: 简单却有效的多轮对抗攻击学习

SEMA: Simple yet Effective Learning for Multi-Turn Jailbreak Attacks

University of Rochester(罗切斯特大学);Microsoft Research(微软研究院)

Mingqian Feng, Xiaodong Liu, Weiwei Yang, Jialin Song, Xuekai Zhu, Chenliang Xu, Jianfeng Gao

AI总结:

SEMA通过多阶段学习实现高效多轮对抗攻击,无需依赖现有策略,提升攻击成功率并增强安全测试能力。

摘要:

多轮对抗攻击捕捉了安全对齐聊天机器人的真实威胁模型,其中单轮攻击只是特殊情况。然而现有方法在探索复杂性和意图漂移方面存在缺陷。我们提出SEMA,一个简单但有效的框架,通过训练多轮攻击者而无需依赖任何现有策略或外部数据。SEMA分为两个阶段。预填充自调谐通过在非拒绝、结构良好的多轮对抗提示上微调,使可使用展开成为可能,这些提示由最小前缀自动生成,从而稳定后续学习。具有意图漂移意识的强化学习训练攻击者以生成有效的多轮对抗提示,同时保持相同的有害目标。我们通过结合意图对齐、合规风险和细节程度的意图漂移意识奖励,将有害意图锚定在多轮对抗攻击中。我们的开环攻击模式避免依赖受害者反馈,统一了单轮和多轮设置,并减少了探索复杂性。在多个数据集、受害者模型和对抗判断中,我们的方法实现了最先进的攻击成功率(ASR),优于所有单轮基线、手动编写和模板驱动的多轮基线,以及我们的SFT(监督微调)和DPO(直接偏好优化)变体。例如,SEMA在AdvBench上三个闭源和开源受害者模型上实现了平均80.1%的ASR@1,超过现有最佳水平33.9%。该方法紧凑、可重复,并能跨目标转移,为大语言模型(LLM)安全提供更强且更真实的压力测试,使自动红队行动能够暴露和定位故障模式。我们的代码可在:https://github.com/fmmarkmq/SEMA获取。

英文摘要:

Multi-turn jailbreaks capture the real threat model for safety-aligned chatbots, where single-turn attacks are merely a special case. Yet existing approaches break under exploration complexity and intent drift. We propose SEMA, a simple yet effective framework that trains a multi-turn attacker without relying on any existing strategies or external data. SEMA comprises two stages. Prefilling self-tuning enables usable rollouts by fine-tuning on non-refusal, well-structured, multi-turn adversarial prompts that are self-generated with a minimal prefix, thereby stabilizing subsequent learning. Reinforcement learning with intent-drift-aware reward trains the attacker to elicit valid multi-turn adversarial prompts while maintaining the same harmful objective. We anchor harmful intent in multi-turn jailbreaks via an intent-drift-aware reward that combines intent alignment, compliance risk, and level of detail. Our open-loop attack regime avoids dependence on victim feedback, unifies single- and multi-turn settings, and reduces exploration complexity. Across multiple datasets, victim models, and jailbreak judges, our method achieves state-of-the-art (SOTA) attack success rates (ASR), outperforming all single-turn baselines, manually scripted and template-driven multi-turn baselines, as well as our SFT (Supervised Fine-Tuning) and DPO (Direct Preference Optimization) variants. For instance, SEMA performs an average 80.1% ASR@1 across three closed-source and open-source victim models on AdvBench, 33.9% over prior SOTA. The approach is compact, reproducible, and transfers across targets, providing a stronger and more realistic stress test for large language model (LLM) safety and enabling automatic redteaming to expose and localize failure modes. Our code is available at: https://github.com/microsoft/SEMA.

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。

↑