安全、治理与可靠性 颠覆级 有讲解视频
发表时间
2026-07-22
arXiv
2607.19829

收录解读

现有大语言模型安全评估与防御方法大多基于静态策略,无法应对不断演化的真实攻击手法,亟需动态、自适应的攻防框架。

提出DARWIN框架,将越狱攻击建模为开放式演化过程,通过策略发现、变异、选择与反馈组合实现攻击演化;同时引入在线对抗训练范式,让防护模型从新兴攻击样本中持续学习,兼顾鲁棒性与实用性。

在多个前沿模型上攻击成功率近100%(DeepSeek-V4-Pro、YuFeng-XGuard),GPT-5.5上超90%;防御侧在12个安全基准上平均不安全召回率91.6%,良性样本通过率近100%,超越现有强防护模型。

为LLM安全评估与防护提供了可复用的演化攻防框架,但尚未公开代码及实际部署验证,长期鲁棒性需进一步评估。

原始摘要

Most existing LLM safety evaluation and defense methods follow a static formulation: jailbreak vulnerabilities are evaluated with fixed attack methods, and guardrails are trained on fixed malicious prompt datasets. However, real-world adversaries continuously evolve their capabilities and expand the attack space. To address this challenge, we propose DARWIN, an evolutionary attack-defense framework that formulates jailbreaking as an open-ended evolution process and continuously updates guardrails through an evolving attack-defense loop. DARWIN-Attack is an evolutionary adversary that expands its capabilities through strategy discovery, mutation, selection, and feedback-driven composition. It collects strategies from broad external sources, generates new variants through self-reflection and genetic evolution, and retains effective strategies based on their performance against aligned LLMs. During attack execution, DARWIN-Attack adaptively selects and combines evolved strategies according to feedback from target LLMs and guardrails. Across frontier models and guardrails, it achieves state-of-the-art attack success rates, including nearly 100% on DeepSeek-V4-Pro and YuFeng-XGuard and over 90% on GPT-5.5. On the defense side, we introduce DARWIN-Guard, an online adversarial training paradigm that iteratively learns from emerging adversarial samples generated by DARWIN-Attack. To improve robustness without sacrificing utility, DARWIN-Guard jointly trains on malicious and benign disguised queries, encouraging the model to identify underlying intent rather than superficial attack patterns. DARWIN-Guard achieves an average unsafe recall of 91.6% across 12 safety benchmarks, outperforming strong guardrails such as YuFeng-XGuard and Nemotron Guard, while maintaining a nearly 100% pass rate on standard benign datasets.

解读视频

相关论文

链接