AI Security AI安全 9h ago Updated 2h ago 更新于 2小时前 46

14 Trojanized npm Packages Drop RedC2 4.0 Linux Backdoor With AI-Assisted C2 14个被篡改的npm包投放搭载AI辅助C2的RedC2 4.0 Linux后门

Fourteen trojanized npm packages disguised as calendar and streak utilities were found to silently deliver RedC2 4.0, an AI-powered Linux backdoor, requiring only a single import to execute the payload RedC2 4.0 features an LLM-driven component called Red Agent that translates natural-language prompts into beacon commands, lowering the barrier for complex post-exploitation operations The malicious packages bundle a Linux beacon (RedShell) disguised as native math accelerator binaries, launching 14个恶意npm包伪装成日历和进度追踪工具,实际在Linux系统中植入RedC2 4.0后门 RedC2 4.0是跨平台C2框架,包含AI驱动的Red Agent组件,可通过自然语言指令执行渗透测试任务 恶意包无需安装钩子,仅需在依赖图中被引用即可自动执行载荷,隐蔽性极强 该框架售价99.99美元,支持Windows/macOS/Linux,具备权限提升、内存执行、横向移动等高级功能 攻击者利用AI降低使用门槛,使不同技能水平的操作者都能高效执行复杂的多阶段入侵

68
Hot 热度
65
Quality 质量
62
Impact 影响力

Analysis 深度分析

TL;DR

  • Fourteen trojanized npm packages disguised as calendar and streak utilities were found to silently deliver RedC2 4.0, an AI-powered Linux backdoor, requiring only a single import to execute the payload
  • RedC2 4.0 features an LLM-driven component called Red Agent that translates natural-language prompts into beacon commands, lowering the barrier for complex post-exploitation operations
  • The malicious packages bundle a Linux beacon (RedShell) disguised as native math accelerator binaries, launching it as a detached background process without needing install hooks or exported functions
  • RedC2 4.0 supports surveillance, credential theft, in-memory execution, lateral movement, SOCKS5 proxying, and network pivoting across Windows, macOS, and Linux
  • This attack follows a coordinated supply chain compromise of Rust crates, signaling a broader trend of AI-integrated C2 frameworks being distributed through package ecosystems

Why It Matters

This incident highlights the growing convergence of AI capabilities with supply chain attacks, where threat actors are embedding LLM-driven command execution into commercial-grade C2 frameworks distributed through legitimate package registries. For AI practitioners and security professionals, it demonstrates how AI-assisted tooling is being weaponized to lower the skill threshold for sophisticated intrusions, making complex multi-stage attacks accessible to operators of varying expertise levels.

Technical Details

  • Trojanized Packages: 14 npm packages (e.g., streak-metrics-math, kit-map-vim) masquerade as functional date/streak utilities while bundling malicious binaries named math-core.bin, calc.bin, etc., located in dist/ or dist/internal/
  • Execution Mechanism: The dist/index.mjs entry file acts as a trojan loader that re-exports legitimate date helpers and immediately launches the bundled implant as a detached background process—no install hook or exported function required; a single transitive import triggers execution
  • RedShell Linux Beacon: Provides interactive /bin/sh access, system discovery, file operations, SSH key and browser credential harvesting, in-memory ELF execution, SOCKS5 proxying, and network pivoting; registers with C2 via check-in messages and enters a command-processing loop
  • Red Agent (LLM Component): An LLM-backed command execution layer that converts natural-language prompts into framework beacon commands, enabling operators to orchestrate network reconnaissance and credential dumping through conversational interfaces
  • Cross-Platform Capabilities: Windows variant includes UAC bypass, AV/EDR tampering, and lateral movement; macOS lacks these but shares core surveillance and data collection features; all three platforms support multi-beacon operation and host-to-host tunneling

Industry Insight

  • Package registry supply chains remain a critical attack vector; developers and organizations should implement strict dependency auditing, SBOM generation, and runtime behavior monitoring to detect anomalous binary execution from seemingly benign packages
  • The integration of LLMs into C2 frameworks represents a paradigm shift in offensive tooling—security teams should anticipate similar AI-augmented attack tools and update detection strategies to identify natural-language-to-command translation patterns in network traffic
  • The $99.99 price point and documented evasion-first design philosophy suggest commercialization of AI-powered red team tools is accelerating; organizations should factor AI-enhanced threat actor capabilities into their red team exercises and threat modeling scenarios

TL;DR

  • 14个恶意npm包伪装成日历和进度追踪工具,实际在Linux系统中植入RedC2 4.0后门
  • RedC2 4.0是跨平台C2框架,包含AI驱动的Red Agent组件,可通过自然语言指令执行渗透测试任务
  • 恶意包无需安装钩子,仅需在依赖图中被引用即可自动执行载荷,隐蔽性极强
  • 该框架售价99.99美元,支持Windows/macOS/Linux,具备权限提升、内存执行、横向移动等高级功能
  • 攻击者利用AI降低使用门槛,使不同技能水平的操作者都能高效执行复杂的多阶段入侵

为什么值得看

本文揭示了AI技术如何被恶意利用来降低网络攻击门槛,对安全从业者和企业供应链安全具有重要警示意义。npm生态系统的恶意包攻击趋势表明,传统依赖检查机制已不足以应对新型威胁,需要建立更全面的软件供应链安全监控体系。

技术解析

  • 恶意包传播机制:14个npm包(如streak-metrics-math、kit-map-vim等)伪装成功能正常的日历/进度追踪工具,实际在dist/index.mjs中嵌入后门加载器,通过重新导出日期辅助函数并自动启动捆绑的Linux后门程序
  • RedC2 4.0架构:跨平台C2框架,包含RedShell Linux信标,支持交互式shell、文件操作、数据收集(SSH密钥、浏览器凭据)、内存中ELF执行、SOCKS5代理和网络 pivoting
  • AI集成能力:Red Agent组件基于LLM,将自然语言指令转换为框架信标命令,支持网络侦察、凭据转储等复杂后渗透任务,大幅降低操作门槛
  • 隐蔽执行技术:无需安装钩子函数,单包引用即可触发载荷执行;后门文件伪装为数学加速器(math-core.bin等),位于dist/或dist/internal/目录下

行业启示

  • 供应链安全威胁升级:AI赋能的恶意软件通过npm等包管理器传播,表明攻击者正利用开源生态的便利性降低攻击成本,企业需加强依赖项安全审计和软件物料清单(SBOM)管理
  • AI双刃剑效应:Red Agent等AI组件使非专业攻击者也能执行复杂渗透,安全团队需建立针对AI辅助攻击的检测能力,同时关注合法红队工具与恶意软件的边界模糊化趋势
  • 防御策略建议:实施严格的npm包来源验证、运行时行为监控和异常进程检测,特别关注无安装钩子的自动执行模式和AI驱动的命令转换行为

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 LLM 大模型 Open Source 开源