Downscaling Attack and Defense: Turning What You See Back Into What You Get

A New Defense Against Adversarial Images: Turning a Weakness into a Strength

Natural images are virtually surrounded by low-density misclassified regions that can be efficiently discovered by gradient-guided search --- enabling the generation of adversarial images. While many techniques for detecting these attacks have been proposed, they are easily bypassed when the adversary has full knowledge of the detection mechanism and adapts the attack strategy accordingly. In this paper, we adopt a novel perspective and regard the omnipresence of adversarial perturbations as a strength rather than a weakness. We postulate that if an image has been tampered with, these adversar…

私有记忆编辑：将记忆转化为防御以增强大语言模型的数据隐私

Private Memorization Editing: Turning Memorization into a Defense to Strengthen Data Privacy in Large Language Models

本文提出私有记忆编辑（PME）方法，通过检测并编辑LLMs对训练数据中PII的记忆，将模型的记忆能力转化为隐私防御策略，在不影响底层语言模型的前提下有效降低隐私训练数据提取攻击的泄露量，甚至使攻击准确率降为零。

CSC：将对手的毒药转为自身武器

CSC: Turning the Adversary's Poison against Itself

本文提出CSC防御机制，通过特征提取和聚类分析识别 poisoned 样本，将其重新标记为虚拟类，并通过交叉熵损失优化模型，有效降低后门攻击成功率，同时保持模型精度。

使生成模型退化：数据投毒攻击的力量

Turning Generative Models Degenerate: The Power of Data Poisoning Attacks

本研究系统探讨了通过前缀微调对大型语言模型进行数据投毒攻击，发现超参数和触发器设计是关键因素，且现有防御无效，为AI安全防御提供参考。

CoDeL：针对基于LLM的智能体中间接提示注入的协同进化防御

CoDeL: Co-Evolutionary Defense against Indirect Prompt Injection in LLM-based Agents

针对LLM智能体的间接提示注入威胁，提出协同进化防御CoDeL，通过动态重塑攻击分布并联合优化安全与任务完成，显著降低攻击成功率。

将隐私保护机制反用于联邦学习

Turning Privacy-preserving Mechanisms against Federated Learning

本文揭示了结合差分隐私与社区驱动方法的联邦学习配置存在关键安全缺陷，并提出一种包含对抗模式和后门模式的双模式攻击，实验证明其能有效欺骗现有防御机制。