AI Security AI安全 7h ago Updated 1h ago 更新于 1小时前 41

UAT-10147 Uses AI to Scale Server Attacks, Deploys SPECTRE With EDR Bypass and Linux Rootkit UAT-10147利用AI规模化服务器攻击,部署SPECTRE配合EDR绕过和Linux根kit

UAT-10147 is a Chinese-speaking cybercrime group leveraging AI-powered tools across the entire attack lifecycle, from reconnaissance and exploitation to persistence and data exfiltration The group deployed a novel cross-platform implant called SPECTRE alongside established malware like BadIIS, Gh0stCringe, and Noodle RAT to target Windows and Linux servers AI tools including PentestGPT and DeepAudit were integrated into offensive operations for automated vulnerability scanning, exploit validatio 中国语网络犯罪组织UAT-10147利用AI工具规模化攻击Windows和Linux Web服务器,目标覆盖教育、媒体、科技和游戏行业 攻击者使用PentestGPT、DeepAudit等AI驱动框架进行漏洞扫描、利用验证和自动化渗透,实现攻击链规模化 新型跨平台植入物SPECTRE配合EDR绕过技术和Linux Rootkit部署,结合BadIIS恶意软件进行SEO欺诈和数据窃取 攻击者利用合法云服务(如Nacos配置管理)作为异步数据外泄通道,规避传统检测机制

62
Hot 热度
58
Quality 质量
55
Impact 影响力

Analysis 深度分析

TL;DR

  • UAT-10147 is a Chinese-speaking cybercrime group leveraging AI-powered tools across the entire attack lifecycle, from reconnaissance and exploitation to persistence and data exfiltration
  • The group deployed a novel cross-platform implant called SPECTRE alongside established malware like BadIIS, Gh0stCringe, and Noodle RAT to target Windows and Linux servers
  • AI tools including PentestGPT and DeepAudit were integrated into offensive operations for automated vulnerability scanning, exploit validation, and post-exploitation workflow automation
  • Attackers exploited known vulnerabilities in Zimbra, AjaxPro, Telerik UI, and Alibaba Nacos, using Nacos instances as asynchronous exfiltration sinks to blend malicious traffic with legitimate operations
  • The campaign targeted approximately 170,000 URLs across education, media, technology, and gaming sectors, with primary victims in Brazil, Bolivia, China, Canada, and Vietnam

Why It Matters

This case represents a significant escalation in how threat actors operationalize AI tools for offensive purposes at scale, moving beyond manual exploitation to automated, AI-assisted attack chains. The integration of autonomous pentesting frameworks and AI-driven vulnerability scanning into criminal tradecraft signals a new paradigm where open-source security tools are being weaponized against the organizations that produced them. For defenders, this underscores the critical importance of patch management, as the group relied almost entirely on publicly disclosed vulnerabilities and freely available exploitation frameworks.

Technical Details

  • AI-Powered Tooling: The group deployed PentestGPT (an autonomous pentesting framework) on their C2 server to scan web servers and execute PoC exploits, and installed DeepAudit (an AI-driven vulnerability scanning framework) for reconnaissance and potential defensive auditing of their own infrastructure
  • Windows Attack Chain: Initial access via known vulnerabilities leading to RCE on IIS servers, followed by deployment of a web shell, privilege escalation using EfsPotato, Microsoft Defender exclusion configuration, and installation of BadIIS, Quasar RAT (hidden under a scheduled task named "Google Chrome Start"), and Gh0stCringe
  • Linux Attack Chain: Initial foothold through known vulnerabilities, followed by systematic Local Privilege Escalation (LPE) using CVE-2022-0995, CVE-2021-3156, CVE-2015-5287, CVE-2015-3246, CVE-2010-3904, and CVE-2022-0847 to achieve root access, then deployment of Noodle RAT, SPECTRE, and Meterpreter backdoors
  • C2 and Exfiltration Infrastructure: Exfiltrated data was routed through legitimate cloud-based Alibaba Nacos configuration management instances, acting as an asynchronous exfiltration sink that allows attackers to poll their own Nacos instance to verify exploitation success without maintaining persistent reverse shells or direct inbound connections
  • ASP.NET ViewState Exploitation: The group utilized an AI-generated guide leveraging the badsecrets library with publicly known or leaked ASP.NET MachineKey configurations, combined with ysoserial.net to construct malicious deserialization payloads that bypass View State protection and achieve remote code execution
  • Malware Ecosystem: Deployed a previously unreported cross-platform implant dubbed SPECTRE, the BadIIS malware variant operating under a malware-as-a-service (MaaS) model, Quasar RAT, and Noodle RAT (a variant combining Gh0st RAT and Rekoobe features)

Industry Insight

  • The weaponization of open-source AI security tools like PentestGPT and DeepAudit by threat actors demonstrates that defensive tooling is a double-edged sword; organizations should assume any publicly available security framework can be repurposed for offensive operations and should monitor for their presence in compromised environments
  • The use of legitimate cloud infrastructure (Alibaba Nacos) as an exfiltration sink represents an evolving operational security technique that blends malicious traffic with normal administrative activity, suggesting defenders should implement behavioral analytics and DNS/HTTP anomaly detection rather than relying solely on known malicious indicators
  • The group's heavy reliance on publicly disclosed CVEs and freely available exploitation frameworks (Metasploit, ysoserial) highlights that basic hygiene—prompt patching of known vulnerabilities, removal of default credentials, and strict network segmentation—remains the most effective defense against this class of automated, scale-driven attacks

TL;DR

  • 中国语网络犯罪组织UAT-10147利用AI工具规模化攻击Windows和Linux Web服务器,目标覆盖教育、媒体、科技和游戏行业
  • 攻击者使用PentestGPT、DeepAudit等AI驱动框架进行漏洞扫描、利用验证和自动化渗透,实现攻击链规模化
  • 新型跨平台植入物SPECTRE配合EDR绕过技术和Linux Rootkit部署,结合BadIIS恶意软件进行SEO欺诈和数据窃取
  • 攻击者利用合法云服务(如Nacos配置管理)作为异步数据外泄通道,规避传统检测机制

为什么值得看

本文揭示了AI工具如何被网络犯罪组织系统化整合到攻击生命周期中,标志着网络攻击从手工操作向自动化、规模化转变的新阶段。对安全从业者而言,理解这种AI赋能的攻击模式有助于提前部署防御策略,应对日益智能化的网络威胁。

技术解析

  • AI工具集成:攻击者使用PentestGPT(开源自主渗透测试框架)和DeepAudit(AI驱动漏洞扫描框架)进行目标侦察和漏洞验证,实现攻击流程自动化
  • 恶意软件部署:部署SPECTRE跨平台植入物、BadIIS(MaaS模式)、Gh0stCringe和Noodle RAT等后门,配合Quasar RAT和Meterpreter建立持久访问
  • 权限提升技术:Windows端使用EfsPotato进行特权提升并配置Microsoft Defender排除项;Linux端利用CVE-2022-0995、CVE-2021-3156等已知LPE漏洞获取root权限
  • 数据外泄策略:将外泄数据路由至合法云配置管理服务(Nacos),伪装成正常管理流量,避免建立持久反向Shell或直接入站连接

行业启示

  • AI双刃剑效应:开源AI渗透测试工具正被攻击者武器化,安全团队需加强对这类工具的监控和异常使用检测
  • 云基础设施风险:攻击者滥用合法云服务作为C2和数据外泄通道,企业需重新评估云服务的访问控制和流量监控策略
  • 漏洞管理紧迫性:攻击者大规模利用已知漏洞(CVE列表),及时修补和漏洞优先级管理成为防御关键

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