AgentDoxx: Agentic Re-identification of Anonymized Text with Web Search
As Large Language Models (LLMs) gain tool use capabilities such as web search, they can retrieve and cross-reference pub
#cs.AI
共收录 1652 条相关情报。
← 返回所有主题Don't Judge an LLM Only by Its Activations: Discovering Suppressed Safety Features via Counterfactual Activation Potential
Mechanistic interpretability has emerged as the primary means to understand safety behavior of LLMs. However, existing t
AutoDP-LLM: Automating Data Pre-processing for Intrusion Detection Systems using Large Language Models
The increasing complexity and scale of modern cyber-attacks demand intelligent and computationally efficient Intrusion D
Reflections and Fragments: Securing LLMs Against Sequential Mosaic Attacks
Self-play red-teaming improves language-model safety by pitting attacker and defender roles against each other in a zero
Grammar-Guided Code Watermarking with Green Temperature
Large language model watermarking embeds detectable statistical signals during decoding, but the resulting changes to to
Who Is Your Agent Serving? Provider-Side Indirect Prompt Injection in Proactive Agents
Proactive personal agents increasingly decide what to recommend, how to personalize advice, and what follow-up assistanc
Blocking at the Boundary: Auditing Long-Horizon Agents against Staged Prompt Injection
Long-horizon agents consume external content, invoke tools, and modify persistent state. Indirect prompt injection can e
VulValidate: Auditing Function-Level Vulnerability Labels with Executable Evidence
基于学习的漏洞检测高度依赖训练标签的质量。现有大规模数据集（BigVul、PrimeVul、DiverseVul 等）通常以「漏洞修复提交」为依据，把被补丁改动过的函数直接标注为「存在漏洞」，但被安全补丁修改并不等于该函数本身存在漏洞，由此
Hidden Risks of Jev: An Empirical Study of Security, Privacy, and Dual Use
该论文针对新兴的 "Jev" 接口开展首次系统性安全、隐私与双用途研究。Jev 能将自然语言问题转化为带类型的答案与概率，具备低延迟、低成本特点，可作为应用决策层，用于请求路由、工具选择等。作者使用官方 Jev API 与本地可训练、可更新
Runtime Authorization of Self-Generated Subgoals in Long-Horizon Tool-Using AI Agents
Long-horizon tool-using AI agents create subgoals, replan, delegate work, and compose sibling results. Per-tool permissi
Invisible Ink, Visible Lies: How Production Watermarking Causes LLMs to Hallucinate
Text watermarking helps identify AI-generated content, but its effect on factual reliability remains underexplored. In t
Not Self-Decidable: LLMs Cannot Draw the Boundary of What an Agent Verifier Can Check
A verifier for an agent faces rules of two kinds: the ones a fixed check can settle and the ones that require a judge. A
Penumbra: Sample-Efficient Adversarial Search for Regulatory Obligations
Agents are entering finance, healthcare and law, sectors where a violation leaves no lexical signature and carries real
Reactivating Alignment: Defending LLMs from Jailbreaks via Intention-Aware Input-Output Matching
Large language models (LLMs) remain vulnerable to jailbreak attacks that conceal harmful intent within complex adversari
Guess My Weight: Profiled Side-Channel Recovery of Floating-Point Neural-Network Weights
Neural-network parameters deployed on embedded devices may be exposed through physical side-channel leakage during infer
COPEX: Benchmarking LLM Robustness to Adversarial Context Across Model Context Protocol Layers
Large language models increasingly mediate tool use in Model Context Protocol (MCP) systems, where adversarial influence
GrayShield: Bit-Level Sanitization for Transformer Model Supply-Chain Security
该论文关注 Transformer 模型供应链中的一个隐蔽威胁面：BERT、Vision Transformer（ViT）等模型的 32 位浮点权重中，尾数最低有效位（LSB）在数值上几乎不影响推理精度，却可被当作隐蔽信道，将恶意载荷编码进
CyTReX: Explainable AI-Based Cybersecurity Threat Reasoning Framework for DER Networks
Distributed Energy Resource (DER) environments rely on network communication protocols to coordinate control commands, m
Language Model Fingerprinting Requires Rethinking Watermark Teachers
LLM fingerprinting via watermark distillation embeds a statistical watermark signal into model weights, enabling model o
Beware EviLLM: Enabling Vulnerability Injection via Large Language Models
该论文研究了 AI 辅助代码生成（LLM 根据自然语言需求生成代码）场景下的一种新型供应链攻击面，作者将其命名为 EviLLM 攻击。此前相关研究多集中在两类不太现实的威胁模型：一是良性场景，即漏洞是 LLM 无意中引入的；二是非常规威胁模