跳到正文- 多模态检索中的对抗性枢纽性Adversarial Hubness in Multi-Modal Retrieval
Tingwei Zhang, Fnu Suya, Rishi Jha 等会议页面 - GraphRAG 受到攻击GraphRAG under Fire
Jiacheng Liang, Yuhui Wang, Changjiang Li 等会议页面 - Hollow-LLM攻击：零知识验证中的计算空洞权重Hollow-LLM Attack: Computationally Trivial Weights in Zero-Knowledge Verification of LLM Inference
Chen Gong, Beijie Liu, Mengyuan Li会议页面 - MetaBreak:通过特殊 token 操纵越狱在线 LLM 服务MetaBreak: Jailbreaking Online LLM Services via Special Token Manipulation
Wentian Zhu, Zhen Xiang, Wei Niu 等会议页面站内导读 - SoK:大语言模型对越狱攻击的鲁棒性SoK: Robustness in Large Language Models against Jailbreak Attacks
Feiyue Xu, Hongsheng Hu, Chaoxiang He 等会议页面 - URLcoat:利用网页搜索能力越狱大语言模型URLcoat: Exploiting Web Search Capability to Jailbreak Large Language Models
Yiheng Sun, Linkang Du, Zhou Su 等会议页面 - WRATH：利用水印鲁棒性实施与水印无关的黑盒失效攻击WRATH: Turning Watermark Robustness Against Itself via a Watermark-Agnostic Black-Box Invalidation Attack
Nan Jiang, Juan Hu, Bangjie Sun 等会议页面
- AI 替我写了论文,我只得到一个漏报:商业 AI 文本检测器的有效性度量AI Wrote My Paper and All I Got Was This False Negative:* Measuring the Efficacy of Commercial AI Text Detectors
Seth Layton, Bernardo B. P. Medeiros, Kevin Butler 等会议页面 - 走出象牙塔：评估并增强真实世界中文地下对抗术语检测Breaking Free from Ivory Tower: Evaluating and Enhancing Real-world Chinese Underground Adversarial Jargon Detection
Zhifan Jiang, Mingxuan Liu, Yue Qin 等会议页面 - 评估文本到图像模型生成儿童性虐待材料的概念过滤防御Evaluating Concept Filtering Defenses against Child Sexual Abuse Material Generation by Text-to-Image Models
Ana-Maria Cretu, Klim Kireev, Amro Abdalla 等会议页面 - MusicShield:生成式 AI 时代对音乐人的保护MusicShield: Protection for Musicians in the Era of Generative AI
非 AI 分会场Syed Irfan Ali Meerza, Jian Liu会议页面 - SoK:量子机器学习对抗鲁棒性的批判性评估SoK: Critical Evaluation of Quantum Machine Learning for Adversarial Robustness
Saeefa Rubaiyet Nowmi, Jesus Lopez, Md Mahmudul Alam Imon 等会议页面 - SoK:大语言模型越狱护栏的评估SoK: Evaluating Jailbreak Guardrails for Large Language Models
Xunguang Wang, Zhenlan Ji, Wenxuan Wang 等会议页面
- LLM 增强图神经网络能否抵御投毒攻击？Are LLM-Enhanced Graph Neural Networks Robust against Poisoning Attacks?
Yuhang Ma, Jie Wang, Zheng Yan会议页面 - Ensemble Conformal Predictor (EnCP):对数据投毒攻击具有鲁棒性保证的共形预测器Ensemble Conformal Predictor (EnCP): A New Conformal Predictor with Robustness Guarantees against Data Poisoning Attacks
Yuxin Yang, Qiang Li, Runyang Feng 等会议页面 - 利用排行榜大规模分发投毒模型Exploiting Leaderboards for Large-Scale Distribution of Poisoned Models
Anshuman Suri, Harsh Chaudhari, Yuefeng Peng 等会议页面 - 谁教会了谎言？检索增强生成中投毒知识的责任归因Who Taught the Lie? Responsibility Attribution for Poisoned Knowledge in Retrieval-Augmented Generation
Baolei Zhang, Haoran Xin, Yuxi Chen 等会议页面 - 你的编译器在给模型植入后门:深度学习编译器的编译不一致漏洞Your Compiler is Backdooring Your Model: Understanding and Exploiting Compilation Inconsistency Vulnerabilities in Deep Learning Compilers
Simin Chen, Jinjun Peng, Yixin He 等会议页面
隐私与数据泄露
11- ARES:基于激活恢复的可扩展联邦学习梯度反演攻击ARES: Scalable and Practical Gradient Inversion Attack in Federated Learning through Activation Recovery
非 AI 分会场Zirui Gong, Leo Yu Zhang, Yanjun Zhang 等会议页面 - 超越不可区分性:度量 LLM API 的训练数据提取风险Beyond Indistinguishability: Measuring Extraction Risk in LLM APIs
非 AI 分会场Ruixuan Liu, David Evans, Li Xiong会议页面 - LLM Unlearning 应当与知识形式无关LLM Unlearning Should Be Form-Independent
Xiaotian Ye, Mengqi Zhang, Shu Wu会议页面 - LLMThief:评估商业 LLM 应用商店中的配置泄露风险LLMThief: Evaluating Configuration Leaking Risks in Commercial LLM App Stores
非 AI 分会场Pinji Chen, Jinlong Jiang, Jianjun Chen 等会议页面 - 联邦学习中主动梯度反演攻击的可检测性On the Detectability of Active Gradient Inversion Attacks in Federated Learning
非 AI 分会场Vincenzo Carletti, Pasquale Foggia, Carlo Mazzocca 等会议页面 - 通过内存取证恢复并重新托管移动端本地大语言模型对话与上下文Recovering and Rehosting Mobile Local LLM Conversations and Contexts via Memory Forensics
Haichuan Xu, David Oygenblik, Runze Zhang 等会议页面 - Shared Spotlight Meridian：可扩展联邦学习的分布式稀疏伪随机函数Shared Spotlight Meridian: Distributed Sparse Pseudorandom Functions for Scalable Federated Learning
非 AI 分会场Youlong Ding, Peihua Mai, Jingqi Zhang 等会议页面 - 声音背后的人:多模态大模型对音频私密属性画像的研究The Person Behind the Sound: Demystifying Audio Private Attribute Profiling via Multimodal Large Language Models
Lixu Wang, Kaixiang Yao, Xinfeng Li 等会议页面 - 面向联邦大语言模型的高效成员推断攻击:投影残差方法Toward Efficient Membership Inference Attacks against Federated Large Language Models: A Projection Residual Approach
非 AI 分会场Guilin Deng, Silong Chen, Yuchuan Luo 等会议页面 - VerfCNN:面向卷积神经网络的最优复杂度 zkSNARKVerfCNN, Optimal Complexity zkSNARK for Convolutional Neural Networks
非 AI 分会场Wenjie Qu, Yanpei Guo, Yue Ying 等会议页面 - 设计师遇上生成式AI：提示到设计生成器在隐私暗黑模式中的作用When Designers Meet GenAI: Understanding the Role of Prompt-to-Design Generators in Privacy Dark Patterns
Jingzhou Ye, Zhaojie Hu, Yao Li 等会议页面
AI 与网络攻防
10- Cottontail：LLM驱动的符号执行测试输入生成Cottontail: Large Language Model-Driven Concolic Execution for Highly Structured Test Input Generation
非 AI 分会场Haoxin Tu, Seongmin Lee, Yuxian Li 等会议页面 - deepSURF：通过大语言模型增强的模糊测试框架检测 Rust 内存安全漏洞deepSURF: Detecting Memory Safety Vulnerabilities in Rust Through Fuzzing LLM-Augmented Harnesses
非 AI 分会场Georgios Androutsopoulos, Antonio Bianchi会议页面 - Incalmo：用于多主机网络红队测试的自主大语言模型辅助系统Incalmo: An Autonomous LLM-assisted System for Red Teaming Multi-Host Networks
Brian Singer, Keane Lucas, Lakshmi Adiga 等会议页面 - 叩击前门：LLM引导的DNS查询处理漏洞系统分析Knocking on the Front Door: An LLM-Guided Systematic Analysis of DNS Query Processing Vulnerabilities
非 AI 分会场Yuqi Qiu, Xiang Li, Zheli Liu会议页面 - 安全运营中心中的大语言模型：人机协作实证研究LLMs in the SOC: An Empirical Study of Human-AI Collaboration in Security Operations Centres
Ronal Singh, Shahroz Tariq, Fatemeh Jalalvand 等会议页面 - PILOT：通过路径引导的迭代式大语言模型提示进行命令行接口模糊测试PILOT: Command-line Interface Fuzzing via Path-Guided, Iterative Large Language Model Prompting
非 AI 分会场Momoko Shiraishi, Yinzhi Cao, Takahiro Shinagawa会议页面 - RadKey:LLM 引导的射频反向散射系统实现穿墙击键推断RadKey: An LLM-Guided RF Backscatter System for Through-Wall Keystroke Inference
非 AI 分会场Qijun Wang, Chunqi Qian, Huacheng Zeng会议页面 - SpecAuditor：为大语言模型驱动的漏洞检测生成审计规范SpecAuditor: Generating Audit Specifications for LLM-Driven Bug Detection
非 AI 分会场Miaoqian Lin, Hao Chen会议页面 - 通过强化学习为文本模糊测试专门化语言模型Specializing Language Models for Textual Fuzzing via Reinforcement Learning
非 AI 分会场Jiayi Lin, Liangcai Su, Junzhe Li 等会议页面 - StepStone:通过用户态库用 LLM 对 GPU 内核驱动做模糊测试StepStone: LLM-Based GPU Kernel Driver Fuzzing via User-Space Libraries
非 AI 分会场Xiaochen Zou, Juefei Pu, Arrdya Srivastav 等会议页面