2026-10-11 归档

今日共收录 1127 条安全情报，包含 372 个 CVE，412 篇安全通告，以及 88 篇研究论文。

学术论文2026-10-08 发布INFO

Characterizing Statistical Separability in TP-CRIV for Probabilistic AI Models

Third-party challenge-response identity verification (TP-CRIV) enables an independent verifier to assess whether a claim

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.AI

学术论文2026-10-08 发布INFO

NOMOS: Compiling Written Policies into Statically Verified Tool-Call Gates for LLM Agents

Tool-using LLM agents violate the policies they are deployed to enforce, often silently. Prior defenses hand-write rules

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.AI #cs.LG

学术论文2026-10-07 发布INFO

Speedbumps: Rejection Attacks on Speculative Decoding

Speculative decoding is a popular technique for increasing the speed and reducing the costs of large language model (LLM

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.AI

学术论文2026-10-07 发布INFO

BRANCH: Bypassing Multi-Scanner AI Guardrails

AI systems increasingly rely on Large Language Models (LLMs) as core reasoning engines, making them targets for prompt i

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.AI

学术论文2026-10-07 发布INFO

DITTO: A Context-aware Pickle-based Pre-Trained Model Scanner for Effective Security Audits

Pre-trained models (PTMs) are widely distributed as serialized binaries, but their reuse often exposes software supply c

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.SE

学术论文2026-10-07 发布INFO

SLDR: Defending Against Malicious Fine-tuning via Selective Layers Recovery and Dynamic Routing

Fine-tuning-as-a-service enables users to adapt aligned large language models (LLMs) to specialized tasks, but malicious

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.AI

学术论文2026-10-07 发布INFO

PatchBench: Measuring Collateral Damage in Activation Patching

An LLM safety patch can pass a benchmark while still being a poor repair. This risk is especially acute for jailbreak re

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.LG #cs.CL

学术论文2026-10-07 发布INFO

On the Reliability of LLM-Based Vulnerability Patching Benchmarks

Large language models (LLMs) have shown strong potential for automated vulnerability patching, but current benchmarks ca

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.SE

学术论文2026-10-07 发布INFO

Constrained-Action AI Remediation for SIEM/XDR via a NeMo-Guardrails Proxy

Security Operations Centers (SOCs) for information technology and operational technology share one incident-response pro

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.AI #cs.CL

学术论文2026-10-07 发布INFO

Formal Runtime Verification for Tool-Using LLM Agents: An Offline Same-Benchmark Study on AgentDojo and STAC

Guardrails for tool-using LLM agents are usually application-specific rules, which makes multi-step, data-dependent safe

历史摘要 · 未重新核验 · arXiv cs.CR

学术论文2026-10-07 发布INFO

Correct Answers, Unsupported Findings: Evidence Binding in Forensic Reconstruction of LLM Agent Logs

Forensic reconstruction of LLM-agent actions requires not only recovering the correct value, but establishing which pres

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.AI

学术论文2026-10-07 发布INFO

MARS: Malware Analysis with Rule-Based Scoring of LLM Claims

Large language models can triage malware through direct verdicts or behavioral claims scored by an external policy. We p

历史摘要 · 未重新核验 · arXiv cs.CR

#cs.AI

安全公告2026-10-09 发布INFO

Vikunja: Explicit lower-permission share on a sub-project is silently overridden by an inherited parent permission (broken access control / privilege-management regression in v2...

## Description ### Summary In Vikunja v2.6.0 the project-permission engine resolves a user's effective permission on a