DiffuGuard: 差分扩散语言模型中内在安全的丧失与恢复
DiffuGuard: How Intrinsic Safety is Lost and Found in Diffusion Large Language Models
AI总结：
本文分析了差分扩散语言模型在对抗攻击中的脆弱性，提出DiffuGuard框架通过随机退火重标记和块级审计修复提升模型安全性，有效降低攻击成功率。
摘要：
差分扩散语言模型（dLLMs）的快速发展引入了与自回归语言模型根本不同的漏洞，源于其迭代和并行生成机制。本文深入分析了dLLM在两种维度上的漏洞：步骤内和步骤间动态。实验结果揭示了标准贪婪重标记策略中的有害偏差，并识别出关键现象——去噪路径依赖，其中早期阶段标记的安全性显著影响最终输出。这些发现表明，尽管当前解码策略构成重大漏洞，dLLMs具有显著的内在安全性潜力。为解锁此潜力，我们提出DiffuGuard，一种无需训练的防御框架，通过双阶段方法：随机退火重标记动态引入受控随机性以缓解贪婪选择偏差，而块级审计与修复利用内部模型表示进行自主风险检测和引导修正。在四个dLLMs上的全面实验表明，DiffuGuard表现出色，将六种不同对抗方法的攻击成功率从47.9%降至14.7%，同时保持模型的实用性和效率。我们的代码可在：https://github.com/niez233/DiffuGuard获取。
英文摘要：
The rapid advancement of Diffusion Large Language Models (dLLMs) introduces unprecedented vulnerabilities that are fundamentally distinct from Autoregressive LLMs, stemming from their iterative and parallel generation mechanisms. In this paper, we conduct an in-depth analysis of dLLM vulnerabilities to jailbreak attacks across two distinct dimensions: intra-step and inter-step dynamics. Experimental results reveal a harmful bias inherent in the standard greedy remasking strategy and identify a critical phenomenon we term Denoising-path Dependence, where the safety of early-stage tokens decisively influences the final output. These findings also indicate that while current decoding strategies constitute a significant vulnerability, dLLMs possess a substantial intrinsic safety potential. To unlock this potential, we propose DiffuGuard, a training-free defense framework that addresses vulnerabilities through a dual-stage approach: Stochastic Annealing Remasking dynamically introduces controlled randomness to mitigate greedy selection bias, while Block-level Audit and Repair exploits internal model representations for autonomous risk detection and guided correction. Comprehensive experiments on four dLLMs demonstrate DiffuGuard's exceptional effectiveness, reducing Attack Success Rate against six diverse jailbreak methods from 47.9% to 14.7% while preserving model utility and efficiency. Our code is available at: https://github.com/niez233/DiffuGuard.
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。
