BiShield-TEE: On the (In-)Security of Unilateral Weight Obfuscation in On-Device TEE-Shielded LLM Partition

On-device deployment of large language models (LLMs) is increasingly adopted due to latency and privacy requirements, bu

历史摘要 · 未重新核验 · arXiv cs.CR

What Survives Self-Evolution: Provenance Marks in Copied Agent Harnesses

Agent harnesses, the prompt and tool code around a base model, leak. Claude Code's 512,000 lines were mirrored within ho

历史摘要 · 未重新核验 · arXiv cs.CR

Evaluating and Improving Static Application Security Testing Tools for Rust

Rust has become a popular systems programming language in recent years because of its strong guarantees on memory and co

历史摘要 · 未重新核验 · arXiv cs.CR

QuasiLeak-Bench: Evaluating LLM-Based Privacy Detectors on Policy-Conditioned Quasi-Evidence Group Recovery

Natural language messages often contain quasi-identifiers that are weak in isolation but can jointly reveal an individua

历史摘要 · 未重新核验 · arXiv cs.CR

Combining LLM and Dynamic Analysis for Universal Testing of Static Application Security Testing Tools

Static application security testing (SAST) tools help detect potential software vulnerabilities, but they themselves con

历史摘要 · 未重新核验 · arXiv cs.CR

Stronger Judges Do Not See More: Deployment-Integrity Blindness in LLM Evaluation

LLM-agent evaluation reports behavioral scores but leaves the identity of the deployed runtime stack unreported. A tampe

历史摘要 · 未重新核验 · arXiv cs.CR

PAGE404: Breaking GPU Isolation Through Hardware-assisted Page Migration

NVIDIA's Grace Hopper (GH200) superchip represents a new generation of heterogeneous computing, tightly integrating a 72

历史摘要 · 未重新核验 · arXiv cs.CR

Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming

We present Screwtape, a pipeline that converts real agent transcripts into attacks that evade action-only monitors. An L

历史摘要 · 未重新核验 · arXiv cs.CR

From Detection to Decision: Calibration and Deferral for Prompt-Injection Detection

Prompt-injection defenses are commonly evaluated as binary detectors, although deployed LLM applications must make opera

历史摘要 · 未重新核验 · arXiv cs.CR

INFUSE: A Unified Surface Benchmark for Indirect Prompt Injection in Agentic Systems

LLM-based agents increasingly execute long-horizon, end-to-end workflows involving shell commands, code modification, de

历史摘要 · 未重新核验 · arXiv cs.CR

From Training to Escalation: Forensic Analysis and Local Recreation of RL/RLHF-Induced Privilege Escalation in AI Agents (CVE-2026-65616)

Mid-2026 reporting on autonomous AI evaluation incidents claimed that frontier large language model (LLM) agents escaped

历史摘要 · 未重新核验 · arXiv cs.CR

P2Patch: Exploit-Guided Vulnerability Repair

Large language models are increasingly deployed as autonomous agents for automated program repair, patching software vul

历史摘要 · 未重新核验 · arXiv cs.CR

Diagnose Before You Refine: Multi-Dimensional Judge Feedback for Query-Efficient Jailbreak Attacks

Automated jailbreak attacks rely on evaluation feedback to optimize adversarial prompts, but existing methods often lack

历史摘要 · 未重新核验 · arXiv cs.CR

lopdf has a stack overflow via deeply nested PDF objects

`lopdf::Document::load_mem` (and the other `load*` entry points) parses nested PDF arrays and dictionaries with unbounde

历史摘要 · 未重新核验 · GitHub Advisory

[原创]CVE-2020-6507利用链

CVE-2020-6507

[原创]CVE-2020-6507利用链

历史摘要 · 未重新核验 · 看雪

CVE-2026-39821：Go idna 纯 ASCII 标签解码漏报

CVE-2026-39821

CVE-2026-39821：Go idna 纯 ASCII 标签解码漏报

历史摘要 · 未重新核验 · 先知社区

CVE-2026-6791：8 MiB 输入击穿 glibc wordexp 栈

CVE-2026-6791

CVE-2026-6791：8 MiB 输入击穿 glibc wordexp 栈

历史摘要 · 未重新核验 · 先知社区

My Credentials, Please: Detecting Filesystem Capability Laundering in MCP Servers at Scale

The Model Context Protocol (MCP) standardizes how AI agents invoke external tools. MCP servers expose tools through a de

历史摘要 · 未重新核验 · arXiv cs.CR

Subgraph Counting with Local Node Differential Privacy

We study the problem of approximately counting subgraphs under local node differential privacy (LNDP), a recently introd

历史摘要 · 未重新核验 · arXiv cs.CR

Language Is a Security Surface for Agent Guards

Step-level guards decide, before each tool call, whether an agent's next action is safe given the user's request and the

历史摘要 · 未重新核验 · arXiv cs.CR