AI Security AI安全 4h ago Updated 2h ago 更新于 2小时前 50

Hackers Start Exploiting Critical Langflow Vulnerability 黑客开始利用Langflow关键漏洞

CVE-2026-0768 is a critical (CVSS 9.8) unauthenticated remote code execution vulnerability in Langflow's custom component editor, caused by insufficient validation of user-supplied strings before Python code execution Threat actors are actively exploiting the vulnerability for reconnaissance and credential harvesting, with attacks primarily originating from Russia and over 360 attempts detected against VulnCheck canaries in the UK All Langflow releases up to version 1.4.2 are affected; the flaw Langflow低代码平台曝出CVSS 9.8分严重RCE漏洞CVE-2026-0768,攻击者可无需认证以root权限执行任意Python代码 漏洞源于自定义组件编辑器中用户输入字符串未经验证即被用于代码执行,存在严重安全缺陷 攻击者正积极利用该漏洞进行系统侦察和凭证收集,主要来源为俄罗斯,已观察到360+次针对英国蜜罐的利用尝试 2026年Langflow漏洞利用活动激增,已发现11个额外漏洞被野外利用,超过15000次攻击成功利用三个已知漏洞

75
Hot 热度
65
Quality 质量
72
Impact 影响力

Analysis 深度分析

TL;DR

  • CVE-2026-0768 is a critical (CVSS 9.8) unauthenticated remote code execution vulnerability in Langflow's custom component editor, caused by insufficient validation of user-supplied strings before Python code execution
  • Threat actors are actively exploiting the vulnerability for reconnaissance and credential harvesting, with attacks primarily originating from Russia and over 360 attempts detected against VulnCheck canaries in the UK
  • All Langflow releases up to version 1.4.2 are affected; the flaw was reported via ZDI in July 2025 and publicly disclosed as a zero-day in January 2026
  • Langflow has become a major attack target in 2026, with 11 additional vulnerabilities now seen exploited in the wild compared to only one prior to 2026
  • Over 15,000 successful attacks have been observed against Langflow instances vulnerable to three other known exploited flaws: CVE-2026-0769, CVE-2025-3248, and CVE-2026-5027

Why It Matters

This vulnerability highlights the growing security risks associated with AI low-code platforms that allow custom code execution, making them attractive targets for threat actors seeking unauthenticated remote code execution. As Langflow adoption increases among AI practitioners, the rapid escalation from a single exploited vulnerability pre-2026 to 12 in 2026 signals a broader trend of attackers prioritizing AI infrastructure supply chains. Organizations deploying Langflow must treat this as an urgent patching priority, especially given the root-level code execution without authentication.

Technical Details

  • Vulnerability: CVE-2026-0768, CVSS 9.8, affecting the code validator in Langflow's custom component editor where user-supplied strings are not properly sanitized before being passed to Python's code execution engine
  • Impact: Unauthenticated attackers can execute arbitrary Python code as root, enabling full system compromise, credential harvesting, and lateral movement
  • Affected versions: All Langflow releases up to and including version 1.4.2
  • Disclosure timeline: Reported through Zero Day Initiative (ZDI) in July 2025; publicly disclosed as a zero-day in January 2026
  • Related exploited vulnerabilities: CVE-2026-0769, CVE-2025-3248, and CVE-2026-5027 — over 15,000 successful exploitation attempts observed across these three flaws
  • Observed attack patterns: Queries for environment variables, secret keys, and SSH access, with exploitation attempts primarily originating from Russia

Industry Insight

  • AI low-code platforms like Langflow are emerging as high-value attack surfaces; organizations should audit all exposed Langflow instances and patch to the latest version immediately, treating unauthenticated RCE as a critical emergency
  • The dramatic increase in Langflow-targeted exploits (from 1 pre-2026 to 12 in 2026) suggests coordinated attacker interest in AI infrastructure — security teams should monitor threat intelligence feeds specifically for AI platform vulnerabilities and implement network-level segmentation for any exposed instances
  • The pattern of credential harvesting and SSH access queries indicates attackers are using Langflow compromises as initial access vectors for broader intrusions; organizations should enforce zero-trust principles, rotate any credentials that may have been exposed on affected systems, and assume compromise on unpatched deployments

TL;DR

  • Langflow低代码平台曝出CVSS 9.8分严重RCE漏洞CVE-2026-0768,攻击者可无需认证以root权限执行任意Python代码
  • 漏洞源于自定义组件编辑器中用户输入字符串未经验证即被用于代码执行,存在严重安全缺陷
  • 攻击者正积极利用该漏洞进行系统侦察和凭证收集,主要来源为俄罗斯,已观察到360+次针对英国蜜罐的利用尝试
  • 2026年Langflow漏洞利用活动激增,已发现11个额外漏洞被野外利用,超过15000次攻击成功利用三个已知漏洞

为什么值得看

Langflow作为流行的AI低代码平台,其安全漏洞直接影响大量AI应用部署和开发流程。该漏洞允许无认证远程代码执行,攻击者可直接获取服务器root权限,对AI基础设施安全构成严重威胁,值得所有使用Langflow的组织和开发者高度关注。

技术解析

  • 漏洞详情:CVE-2026-0768,CVSS评分9.8,存在于Langflow自定义组件编辑器的代码验证器中,用户提供的字符串未经验证即被用于Python代码执行
  • 影响范围:所有Langflow 1.4.2及之前版本均受影响,漏洞于2025年7月通过ZDI报告,2026年1月作为零日漏洞公开
  • 攻击模式:攻击者利用漏洞执行侦察操作,查询环境变量、密钥和SSH访问权限,已成功利用的三个漏洞为CVE-2026-0769、CVE-2025-3248和CVE-2026-5027
  • 利用规模:VulnCheck观察到超过15000次攻击成功利用Langflow实例中的已知漏洞,攻击活动主要来自俄罗斯地区

行业启示

  • AI低代码平台的安全问题日益突出,开发者需要在便捷性和安全性之间取得平衡,重视输入验证和代码执行的安全控制
  • 2026年Langflow漏洞利用活动激增表明攻击者对AI工具链的兴趣持续增长,组织应加强AI基础设施的安全监控和漏洞管理
  • 建议所有使用Langflow的组织立即升级到最新版本,审查自定义组件代码,并实施网络层的安全防护措施以减少攻击面

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。

Open Source 开源 Security 安全 LLM 大模型