Continuing Scans for swagger.json, (Wed, Jun 3rd)
Enterprise applications often still use complex standards like SOAP for web services. The big advantage of SOAP is its t
2026-06-04 归档
今日共收录 1001 条安全情报，包含 200 个 CVE，398 篇安全通告，以及 115 篇研究论文。
← 返回归档列表Chinese hackers use new Atlas RAT malware in European cyberattacks
A Chinese-speaking cybercrime group has expanded its targeting to the European space, deploying previously undocumented
CISA warns of cyberattacks targeting fuel tank monitoring systems
CISA, the FBI, the NSA, the Department of Energy, and other US government partners are warning that hackers are targetin
New 'HTTP/2 Bomb' DoS attack crashes web servers in under a minute
A new denial-of-service (DoS) attack dubbed HTTP/2 Bomb can be launched from a single machine to take down web servers w
CISA warns of active attacks exploiting Android, Linux bugs
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) is warning that hackers are exploiting vulnerabilities
What 345 Days of Untested Exposure Looks Like at a Bank
A two-week penetration test can leave roughly 345 days of real-world exposure unvalidated. Sprocket Security explores wh
Police dismantles 9 crime groups in illegal streaming crackdown
European and international law enforcement agencies have dismantled nine organized crime groups and arrested 29 suspects
Google adds Android protection against AI deepfake scam calls
Google is introducing a new Android security feature that will detect and flag phone calls in which scammers use artific
VS Code zero-day lets hackers steal GitHub tokens in one click
A security researcher has released exploit code for a Visual Studio Code (VS Code) zero-day vulnerability that allows at
Shrinking the IAM Attack Surface through Identity Visibility and Intelligence Platforms (IVIP)
The Fragmented State of Modern Enterprise Identity Enterprise IAM is approaching a breaking point. As organizations sca
New HTTP/2 Bomb Vulnerability Allows Remote DoS on NGINX, Apache, IIS, Envoy & Cloudflare
Cybersecurity researchers have discovered a remote denial-of-service exploit that affects major web servers, including N
NLLog: Lightweight, Explainable SOC Anomaly Detection via Log-to-Language Rewriting
系统日志是安全监控的核心数据源，但传统日志的模板化格式既阻碍自动化分析也不利于人类理解。本文提出 NLLog (Natural-Language Log)，一种轻量级管道，将解析后的日志模板通过确定性规则重写为“谁-做了什么-严重性”的自然
AI Agents Enable Adaptive Computer Worms
该论文提出了一种基于人工智能代理的新型自适应蠕虫威胁。传统计算机蠕虫（如 WannaCry）依赖固定的预定义漏洞进行传播，一旦漏洞被修补即可阻断。而本文展示的AI蠕虫能够针对每个遭遇的目标生成量身定制的攻击策略，通过寄生在被入侵机器上运行开
Which Defense Closes Which Threat? Attributing OWASP-LLM-Top-10 Coverage and Its Brittleness Under Paraphrasing
本研究针对大型语言模型（LLM）应用中的多层级防御组合，提出了一种归因分析方法，以量化每种防御家族具体阻断哪些OWASP LLM Top 10威胁。现有基准测试仅报告单一聚合覆盖率，无法区分不同防御组件的贡献。本文在21个攻击代理基线基础上
From Agent Traces to Trust: Evidence Tracing and Execution Provenance in LLM Agents
本文是一篇系统性综述，聚焦于基于大型语言模型（LLM）的智能体（Agent）在执行过程中的证据追踪与执行溯源问题。随着LLM Agent越来越多地通过外部工具、检索系统、记忆模块、环境以及其他Agent交互来解决复杂任务，其自主能力增强的同
Sequential Data Poisoning in LLM Post-Training
本文研究了LLM后训练阶段（包括监督微调SFT和基于人类反馈的强化学习RLHF或直接偏好优化DPO）中的数据投毒攻击问题。现有工作通常假设每个训练阶段可能发生单次投毒，但忽略了多个攻击者协同的可能性。作者提出了序列化数据投毒威胁模型：多个对
Description-Code Inconsistency in Real-world MCP Servers: Measurement, Detection, and Security Implications
模型上下文协议（MCP）已成为大型语言模型（LLM）利用外部工具的关键标准。在该生态系统中，LLM依赖MCP服务器提供的自然语言描述来选择和执行函数，但这一交互隐式假设工具描述忠实地反映了其实现，而该假设在实践中并未强制验证。因此，MCP部
TIBlender: Early-Warning Threat Intelligence from Cross-Platform Social Media Evidence
该论文提出 TIBlender，一个基于多智能体的大语言模型（LLM）系统，旨在自动化整合来自多个社交媒体平台（X、Reddit、Telegram 和 Discord）的碎片化威胁信号，生成结构化的威胁情报（TI）报告。TIBlender
What If Prompt Injection Never Left? Exploring Cross-Session Stored Prompt Injection in Agentic Systems
本文研究了现代智能体系统（agentic systems）中的跨会话存储提示注入（cross-session stored prompt injection）威胁。传统的大型语言模型（LLM）应用通常在单次会话内与用户交互，且会话结束后状态
When Autoregressive Consistency Hurts Safety Alignment
本文研究了大语言模型（LLM）安全对齐的脆弱性，发现其根源在于对齐效果往往较浅——微调主要重塑模型前几个输出标记的行为。作者提出自回归一致性（autoregressive consistency）概念，即下一个标记预测倾向于保持并延续当前响