AI Security AI安全 7h ago Updated 1h ago 更新于 1小时前 47

Weekly Recap: AI-Powered PLC Attacks, GitLab Attacks, Stripe Key Leaks and More 每周回顾:AI驱动的PLC攻击、GitLab攻击、Stripe密钥泄露等

U.S. agencies warn of active AI-powered attacks targeting internet-exposed Siemens S7 Series PLCs in critical infrastructure, with threat actors using AI-generated scripts and scanning services like Censys and ZoomEye CVE-2026-19478, a critical code injection flaw in GitLab (CVSS 9.4), is under active exploitation allowing unauthenticated modification or deletion of public projects 14 trojanized npm packages were discovered delivering RedC2 4.0, an AI-powered cross-platform backdoor marketed as 美国警告AI驱动的PLC攻击已成为活跃威胁,攻击者利用AI生成漏洞利用脚本针对西门子S7系列PLC,通过Censys/ZoomEye扫描暴露目标 GitLab曝出CVE-2026-19478高危代码注入漏洞(CVSS 9.4),无需认证即可修改或删除公开项目,披露数天内即遭主动利用 14个恶意npm包伪装成日历和习惯追踪工具,实际传播AI驱动的RedC2 4.0跨平台后门,具备监控、凭证窃取和载荷加载能力 学术研究者展示Zombie Card攻击,通过智能手机中继修改CDCVM标志绕过加密检查,可使用过期Visa卡完成非接触支付 俄罗斯黑客组织UNC6293/UNC7005/UNC5976利用

68
Hot 热度
62
Quality 质量
70
Impact 影响力

Analysis 深度分析

TL;DR

  • U.S. agencies warn of active AI-powered attacks targeting internet-exposed Siemens S7 Series PLCs in critical infrastructure, with threat actors using AI-generated scripts and scanning services like Censys and ZoomEye
  • CVE-2026-19478, a critical code injection flaw in GitLab (CVSS 9.4), is under active exploitation allowing unauthenticated modification or deletion of public projects
  • 14 trojanized npm packages were discovered delivering RedC2 4.0, an AI-powered cross-platform backdoor marketed as a C2 framework built for evasion
  • Researchers demonstrated a "Zombie Card" attack that bypasses cryptographic checks to enable contactless payments with physically expired Visa cards via smartphone relay
  • A remote Spectre side-channel attack against Cloudflare Workers can leak JWTs at 12 bits per second, a 360x improvement over 2021 demonstrations

Why It Matters

This week's threats illustrate how AI is lowering the barrier to sophisticated attacks across both OT and IT environments, making exploit development faster and more accessible to threat actors. The convergence of supply chain compromises (npm packages, MSP hijacking), actively exploited zero-days, and hardware-level side-channel attacks demonstrates that defense-in-depth is no longer optional—each layer must be independently secured.

Technical Details

  • Siemens PLC Exploitation Chain: Threat actors use Censys/ZoomEye for reconnaissance, deploy AI-generated scripts disguised as monitoring tools, and leverage read access to map target environments in preparation for write operations that could disrupt industrial processes
  • CVE-2026-19478 (GitLab): A CVSS 9.4 unauthenticated code injection vulnerability requiring no user interaction or special configuration, enabling modification or deletion of publicly accessible GitLab projects
  • RedC2 4.0 Distribution: 14 npm packages disguised as calendar and streak utilities serve as droppers for an AI-powered implant offering surveillance, credential theft, payload loading, and mass-operation capabilities across Windows, macOS, and Linux
  • Zombie Card Attack: Exploits the Consumer Device Cardholder Verification Method (CDCVM) flag via a smartphone relay setup to alter the expiration date presented to PoS terminals without breaking underlying card cryptography
  • Cloudflare Workers Spectre Attack: Uses amplification techniques combined with a remote timing server to achieve 120 bits/hour JWT leakage from co-located Workers in shared production environments

Industry Insight

  • Organizations running Siemens S7 Series PLCs on the internet or with insufficient network segmentation should treat this as an immediate remediation priority, not a theoretical risk—air-gapping and network segmentation are critical
  • The rapid exploitation of GitLab's CVE-2026-19478 underscores the need for automated vulnerability scanning and timely patching pipelines, especially for publicly accessible repositories
  • Supply chain security must extend beyond code repositories to include MSP relationships, captive Wi-Fi infrastructure, and npm dependency audits, as attackers are increasingly targeting these trust boundaries

TL;DR

  • 美国警告AI驱动的PLC攻击已成为活跃威胁,攻击者利用AI生成漏洞利用脚本针对西门子S7系列PLC,通过Censys/ZoomEye扫描暴露目标
  • GitLab曝出CVE-2026-19478高危代码注入漏洞(CVSS 9.4),无需认证即可修改或删除公开项目,披露数天内即遭主动利用
  • 14个恶意npm包伪装成日历和习惯追踪工具,实际传播AI驱动的RedC2 4.0跨平台后门,具备监控、凭证窃取和载荷加载能力
  • 学术研究者展示Zombie Card攻击,通过智能手机中继修改CDCVM标志绕过加密检查,可使用过期Visa卡完成非接触支付
  • 俄罗斯黑客组织UNC6293/UNC7005/UNC5976利用合法认证流程针对欧美学术界、航空航天、政府和智库进行定向网络间谍活动
  • Cloudflare Workers遭遇远程Spectre侧信道攻击,JWT泄露速率达12比特/秒(较2021年提升360倍),每小时可泄露120比特
  • Cl0p勒索软件组织利用PTC Windchill关键漏洞部署定制JSP Web Shell,具备密钥库解密、凭证映射和自定义Java类加载器功能

为什么值得看

本文揭示了AI技术正被攻击者广泛用于降低漏洞利用门槛、增强恶意软件能力,同时供应链攻击和侧信道攻击技术持续演进,对关键基础设施和云原生架构构成现实威胁。

技术解析

  • AI驱动的PLC攻击:攻击者使用Censys和ZoomEye等合法扫描服务发现暴露的西门子S7系列PLC,部署伪装成监控工具的AI生成脚本进行漏洞利用,先获取读访问权限了解目标环境,为后续写入操作造成破坏做准备。
  • GitLab CVE-2026-19478:CVSS 9.4分的代码注入漏洞,允许未认证攻击者在无需凭据、用户交互或特殊配置的情况下修改或删除公开GitLab项目数据,属于高危远程代码执行类漏洞。
  • RedC2 4.0后门:14个npm包伪装成功能性日历和习惯追踪工具,实际植入AI驱动的跨平台C2框架,支持Windows/macOS/Linux,具备隐蔽性、凭证窃取、载荷加载和大规模操作能力,由威胁行为者"MarlboroMan"在Hack Forums推广。
  • Zombie Card攻击:利用智能手机中继设置修改POS终端接收的过期日期,通过调整Consumer Device Cardholder Verification Method (CDCVM)标志绕过加密验证,在多数测试银行成功完成实体商店交易。
  • Cloudflare Workers Spectre攻击:使用放大技术结合远程计时服务器,在共置Worker间实现远程侧信道攻击,JWT泄露速率达120比特/小时,对边缘计算环境的多租户隔离构成严重威胁。
  • Cl0p定制Web Shell:针对PTC Windchill和FlexPLM的JavaServer Pages后门,集成完整勒索平台功能,包括敏感Vault数据映射、Windchill密钥库凭证解密和自定义Java类加载器执行额外代码。

行业启示

  • AI双刃剑效应加剧:攻击者正将AI用于自动化漏洞利用脚本生成和恶意软件增强,安全防御方需同步提升AI驱动的威胁检测和响应能力,缩小攻防技术差距。
  • 供应链攻击持续升级:从npm包投毒到MSP供应链劫持,攻击者越来越倾向于通过可信渠道渗透目标,组织需加强第三方依赖审查和软件物料清单(SBOM)管理。
  • 边缘计算安全需重新评估:Cloudflare Workers侧信道攻击表明多租户边缘环境存在严重隔离风险,云服务商和开发者需重新审视共享基础设施的安全边界设计。

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 Research 科学研究