AI Security AI安全 2d ago Updated 2d ago 更新于 2天前 46

CISA Adds 4 Actively Exploited Adobe, Joomla, and Langflow Flaws to KEV CISA将4个被主动利用的Adobe、Joomla和Langflow漏洞添加到KEV目录

CISA added four actively exploited vulnerabilities to its KEV catalog: two critical Adobe ColdFusion and JoomShaper flaws, one Joomlack Page Builder RCE, and one Langflow authorization bypass. Langflow is under active attack via CVE-2026-55255 (IDOR) combined with CVE-2026-33017 (RCE) to steal LLM and AWS credentials for financial gain. Attackers are leveraging these CMS and framework flaws to deploy web shells, create superuser accounts, and execute arbitrary code with minimal delay after discl CISA将Adobe ColdFusion、Joomla插件及Langflow的4个漏洞列入已知被利用漏洞目录,证实存在活跃攻击。 Langflow漏洞CVE-2026-55255被用于跨租户窃取LLM提供商密钥和AWS凭证,结合RCE漏洞形成完整攻击链。 攻击者利用Joomla插件上传Webshell并创建超级用户,部分漏洞在披露后数小时内即遭利用。 联邦机构需在2026年7月10日前修补相关漏洞,以防范针对AI编排平台和CMS系统的针对性攻击。

65
Hot 热度
70
Quality 质量
60
Impact 影响力

Analysis 深度分析

TL;DR

  • CISA added four actively exploited vulnerabilities to its KEV catalog: two critical Adobe ColdFusion and JoomShaper flaws, one Joomlack Page Builder RCE, and one Langflow authorization bypass.
  • Langflow is under active attack via CVE-2026-55255 (IDOR) combined with CVE-2026-33017 (RCE) to steal LLM and AWS credentials for financial gain.
  • Attackers are leveraging these CMS and framework flaws to deploy web shells, create superuser accounts, and execute arbitrary code with minimal delay after disclosure.
  • Federal agencies must patch affected systems by July 10, 2026, highlighting the urgent need for rapid remediation in exposed enterprise environments.

Why It Matters

This update underscores the critical risk facing organizations using popular CMS platforms and AI orchestration tools like Langflow, which are increasingly targeted for credential theft and remote code execution. For AI practitioners and security teams, it highlights that AI infrastructure is no longer just a target for data exfiltration but a vector for broader system compromise and financial crime. Immediate patching and monitoring for unauthorized file uploads or suspicious API calls are essential to prevent exploitation.

Technical Details

  • Adobe ColdFusion (CVE-2026-48282): A path traversal vulnerability with a CVSS score of 10.0 allowing arbitrary code execution in the context of the current user. Exploitation was observed within hours of disclosure.
  • Joomlack Page Builder (CVE-2026-56290): An improper access control flaw (CVSS 10.0) enabling unauthenticated remote code execution via arbitrary file upload. Attackers used this to deploy web shells, such as bhup.php, in various directories.
  • Langflow Authorization Bypass (CVE-2026-55255): An IDOR vulnerability (CVSS 6.1) allowing authenticated attackers to execute flows belonging to other users by manipulating user-controlled keys. This was used to steal LLM provider keys and AWS credentials.
  • JoomShaper SP Page Builder (CVE-2026-48908): An unrestricted file upload vulnerability (CVSS 10.0) allowing unauthenticated PHP code execution. Attackers uploaded malicious files via specific endpoints to gain superuser access.
  • Attack Chain: Threat actors combined Langflow’s IDOR and RCE flaws to perform reconnaissance, enumerate flows, steal credentials, and deploy second-stage downloaders for botnet or cryptojacking purposes.

Industry Insight

  • Prioritize AI Infrastructure Security: AI orchestration platforms like Langflow contain high-value credentials (API keys, cloud tokens). Treat them with the same security rigor as traditional databases or identity providers.
  • Rapid Patching is Non-Negotiable: The observation of exploitation within hours of disclosure for Adobe and Joomla flaws indicates that automated scanning tools are aggressively targeting known CVEs. Organizations must implement automated vulnerability management and patching pipelines.
  • Monitor for Credential Theft: Given the focus on stealing LLM and cloud keys, implement strict access controls, rotate credentials regularly, and monitor for unusual API usage patterns or unauthorized flow executions in AI platforms.

TL;DR

  • CISA将Adobe ColdFusion、Joomla插件及Langflow的4个漏洞列入已知被利用漏洞目录,证实存在活跃攻击。
  • Langflow漏洞CVE-2026-55255被用于跨租户窃取LLM提供商密钥和AWS凭证,结合RCE漏洞形成完整攻击链。
  • 攻击者利用Joomla插件上传Webshell并创建超级用户,部分漏洞在披露后数小时内即遭利用。
  • 联邦机构需在2026年7月10日前修补相关漏洞,以防范针对AI编排平台和CMS系统的针对性攻击。

为什么值得看

本文揭示了AI基础设施(如Langflow)正成为高级持续性威胁的高价值目标,攻击者不仅追求系统控制权,更旨在窃取关键的API密钥和云凭证。对于AI从业者和安全团队而言,这强调了在部署LLM编排工具时实施严格身份验证、最小权限原则及持续监控的紧迫性。

技术解析

  • Langflow IDOR与RCE组合攻击:CVE-2026-55255是一个授权绕过漏洞(IDOR),允许认证用户通过指定受害者Flow ID执行其他用户的流程;结合CVE-2026-33017(未认证RCE),攻击者可远程执行代码并部署第二阶段下载器,用于窃取LLM和AWS密钥。
  • Adobe ColdFusion路径遍历:CVE-2026-48282(CVSS 10.0)为严重路径遍历漏洞,可导致当前用户上下文下的任意代码执行,公开披露后数小时内即发现来自印度的利用尝试。
  • Joomla插件文件上传漏洞:CVE-2026-56290和CVE-2026-48908均涉及不安全的文件上传,允许未认证用户上传PHP Webshell或危险文件,进而实现远程代码执行和创建超级用户账户。
  • 攻击战术与取证:攻击者表现出高度组织性,先进行身份验证侦察,再枚举Flow,最后利用IDOR和RCE循环执行恶意负载;Webshell通常隐藏在/media或/images等非标准目录下。

行业启示

  • AI平台安全需提升至核心地位:Langflow等AI编排平台因集中管理敏感凭证,已成为黑客眼中的“金矿”,开发者和运维人员必须将其视为高价值资产进行加固。
  • 供应链与第三方组件风险加剧:Adobe和Joomla生态系统的漏洞表明,广泛使用的商业和开源软件仍是攻击入口,企业应建立快速的漏洞响应机制,优先修补CVSS评分高且已被利用的组件。
  • 主动防御与监控至关重要:鉴于漏洞披露后极短的利用窗口期,仅依赖补丁更新不够,需结合行为分析、异常流量检测和定期安全扫描,以识别如IDOR和隐蔽Webshell等高级攻击迹象。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Security 安全