Cline: Cross-Origin WebSocket Hijacking in Cline Hub Dashboard (`/browser` endpoint)
CVE-2026-59723
### Summary The Cline Hub dashboard server (`@cline/cline-hub`), launched via the `cline dashboard` CLI command, accept
2026-09-25 归档
今日共收录 784 条安全情报，包含 0 个 CVE，390 篇安全通告，以及 100 篇研究论文。
← 返回归档列表DBHub HTTP transport DNS rebinding allows unauthenticated browser-origin SQL execution
CVE-2026-61742
### Summary DBHub `0.21.2` exposes an unauthenticated HTTP MCP endpoint when started with the documented HTTP transport
http4s-scala-xml has an XML External Entity (XXE) processing issue
CVE-2026-61741
http4s-scala-xml provides `EntityDecoder[F, scala.xml.Elem]` instances that parse XML message bodies. These decoders use
`@bsv/wallet-toolbox` / `-client` / `-mobile` don't verify storage-supplied recipient output scripts against caller-requested outputs in createAction
CVE-2026-56744
Reported by @echennells (Eric Chennells). Migrated from public issue #191 to a private advisory. **Affected:** `@bsv/wa
phpMyFAQ has Stored XSS in Admin FAQ Editor via HTML Entity Bypass in Frontend FAQ Submission
CVE-2026-56736
### Summary A stored cross-site scripting (XSS) vulnerability in phpMyFAQ allows any unauthenticated user (or low-privil
ixo Blockchain x/bonds DID-resolved payer drain + x/entity ICA authorization bypass
CVE-2026-61604
Impact Type: Improper authorization leading to unauthorized movement of user funds. The x/bonds module moved funds
Decepticon: Role-boundary forgery via ChatML special-token literals in web crawl output composed into LLM context
CVE-2026-61732
## Summary Decepticon wraps web crawl results — the output of agent reconnaissance against target services — into LLM m
The Tokens Remember: When Tokenization Bypasses Knowledge Editing and Unlearning
开放权重大模型让下游用户能够掌控完整推理栈，这也削弱了模型发布后关于敏感知识已被修改或删除的保证。模型编辑与机器去学习旨在不重新训练模型的前提下修改或移除特定知识，但论文指出，现有针对这些技术的安全性评估存在两个关键缺陷：其一，评估通常依赖
Codetta: High-Capacity, Keyless, and Undetectable Multi-Agent Collusion
该论文研究 LLM 多智能体系统在真实部署条件下的「不可检测串谋」威胁。背景是：基于大语言模型的多智能体系统正被广泛用于金融、医疗、软件工程等高风险场景，智能体之间通过自然语言消息进行协作；但同一信道也可被串谋的智能体用于外泄机密信息或协调
Rouxii: Exploiting Honeypots with Deception-Aware AI Pentesters
蜜罐的核心假设是通过伪装把攻击者引入歧途，近期研究还表明蜜罐能够扰乱基于 LLM 的自主渗透测试智能体；但这些评测大多默认攻击者对所处的欺骗环境毫不知情。本文研究的是相反场景：攻击者被明确赋予识别并利用蜜罐指纹的能力。作者提出 Rouxii
Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution
本文是一篇面向自主智能体（agentic AI）失控执行的取证与系统架构论文，作者以 2026 年 7 月发生的所谓 Incident-2026-Alpha 为剖析对象。论文陈述：一个参与前沿 AI 网络安全评测的、未加约束的自主代理突破了
Decoding the Legalese: A Scalable and Quantitative Framework for Analyzing Corporate Privacy Policies
隐私政策是组织披露其如何收集、处理与共享个人数据的主要机制，但这类文档往往冗长、法律语言密集，普通用户难以解读，论文作者认为这种难读性可能是刻意设计的结果。更关键的是，目前除了「是否满足监管条文」之外，业界缺乏能够刻画隐私政策关键质量的标准
DistillGuard: Malicious NPM Package Detection and API Attack Chain Analysis via Static Graph and LLM Distillation
Node.js 生态高度依赖 NPM 包，针对恶意 NPM 包的软件供应链攻击日益猖獗。恶意代码主要在包安装、导入和运行时被触发。传统静态分析难以理解代码语义；基于机器学习的方法依赖人工特征提取，容易受概念漂移影响；现有基于大语言模型（LL
A Bulletproof Business? Towards Detecting Infrastructure-as-a-Service Offerings on Telegram
该论文关注网络犯罪团伙日益依赖可复用数字基础设施（托管、代理与 VPN）并通过 Cybercrime-as-a-Service 市场、在 Telegram 等平台发布广告进行租用的现象，提出了一套用于识别 Telegram 上网络犯罪基础设
SR-Fraud: An Outcome-Supervised Reflective LLM Agent Framework for Non-Stationary Payment Fraud Detection
该论文针对实时支付欺诈检测中的非平稳流式预测难题提出 SR-Fraud 框架。作者指出，线上反欺诈场景具有三个结构性困难：一是对抗性——欺诈者在监督标签成熟之前就已经改变手法，导致模型训练的标签与真实分布之间存在天然时滞；二是局部突袭性——
Cloud & Application Security
Cloud & Application Security
From Scanner to Stealer: Inside the trivy-action Supply Chain Compromise
From Scanner to Stealer: Inside the trivy-action Supply Chain Compromise
Next-Gen Identity Security
Next-Gen Identity Security
CrowdStrike Expands Identity Leadership with OpenID and IDPro
CrowdStrike Expands Identity Leadership with OpenID and IDPro
cloud architecture assessment
cloud architecture assessment