UAT-10147
Summary
Cisco Talos tracks UAT-10147 as a financially motivated, Chinese-speaking intrusion actor that targets internet-exposed Windows and Linux web servers globally, monetizing compromises through SEO fraud and data theft. The actor was first identified in early 2026 after Talos observed a compromised host communicating with a download server (139.180.197[.]150) whose open directory listing exposed the campaign — including a target list of roughly 170,000 URLs. On August 20, 2026 Talos published two reports: one on the campaign's agentic-AI integration and attack chains, and one on the new cross-platform SPECTRE backdoor and the Specter Linux kernel rootkit it carries.
Talos assesses with moderate-to-high confidence that UAT-10147 is part of an emerging class of financially motivated intrusion operators that use agentic AI to operationalize offensive tradecraft at scale — iterative exploit refinement, adaptive troubleshooting, post-exploitation automation, exploit-validation workflows, and operational-documentation generation — marking a transition from AI-assisted scripting toward semi-autonomous offensive orchestration.
Tags
- groups
- actors
- UAT-10147
- Chinese-speaking
- cybercrime
- financially motivated
- web server
- IIS
- SEO fraud
- BadIIS
- SPECTRE
- Specter
- agentic AI
- data theft
Why this matters
- This is a web-server intrusion actor, not a supply-chain actor: initial access is public RCEs against IIS / Nginx/Apache deployments, monetized via SEO-fraud modules (BadIIS MaaS, ASHX
SeoEngineHandler) and credential / data theft. Defenders should read it through patching and IIS hardening, not package-registry controls. - The agentic-AI angle is operationally concrete: recovered artifacts include an AI-generated ASP.NET ViewState RCE guide and four companion Python scripts (check_paths, deploy_implant, deploy_shell, exfil) plus a live findings log confirming RCE on a real target. The durable pattern — machine-key exposure driving a fully scripted, validated exploit loop — is generalizable to any environment that exposes
__VIEWSTATE. - SPECTRE / Specter adds durable technical tradecraft: a Linux kernel rootkit masquerading as
acpi_pad.kowith aBefore=sysinit.targetsystemd unit, ftrace IPMODIFY syscall hooks, and a Windows BYOVD EDR-callback unlink. These are reusable techniques, not just this actor's tools.
Reported activity
- Early 2026: Talos discovers a Chinese-speaking cybercrime group targeting vulnerable web servers; the actor engages in SEO fraud and data theft.
- August 20, 2026: Talos publishes the agentic-AI campaign report and the SPECTRE implant report (both by Joey Chen). Affected servers span Brazil, Bolivia, China, Canada, and Vietnam across government, education, media, technology, and gaming sectors.
- Discovered via OPSEC failure: the actor's C2 download server
139.180.197[.]150had an open directory listing; the target list (~170,000 URLs in 17 files of ~10,000, "w" = 萬) and AI-generated tooling were recovered from it.
Tools and infrastructure
- SPECTRE cross-platform C backdoor (Windows 45-command variant; Linux 29-command variant) with Specter kernel rootkit (
acpi_pad.ko,hardware-monitor.service, ftrace IPMODIFY, magic PID 31337). - BadIIS malware-as-a-service SEO-fraud modules; C# ASHX
SeoEngineHandlerSEO engine (C2 onvn[.]xyz, targets the Cốc Cốc crawler). - Potato family LPE: GodPotato, JuicyPotato, and custom-compiled EfsPotato / RustPotato (PDBs reference
...\Desktop\AI\build paths). - Commodity persistence: QuasarRAT (campaign ID with a derogatory Vietnamese string), Noodle RAT (Type
0x03A2ELF), Meterpreter, Gh0stCringe (shellcode in a Go loader), and a two-layer web shell authenticating on theX-IDheader tokenx9. - AI offensive frameworks on the management/C2 host: DeepAudit, PentestGPT, ysoserial; Metasploit for RCE; exfil via
webhook.siteand a Nacos config-server async sink. - Staging:
139.180.197[.]150(open dir; secondary web-shell staging on port 54321) andadminapi.tippusoni[.]in(BadIISdll.zip/user.txt). - Exploited vulnerabilities: CVE-2022-27925 (Zimbra), CVE-2021-23758 (AjaxPro), CVE-2021-29441 / -29442 (Nacos), CVE-2019-18935 (Telerik), plus LPEs CVE-2022-0995, CVE-2021-3156, CVE-2015-5287, CVE-2015-3246, CVE-2010-3904, CVE-2022-0847.
Defender notes
- Treat any
acpi_pad.koon a host without operator-controlled ACPI thermal management, ahardware-monitor.serviceorderedBefore=sysinit.target, orkill()to PID 31337 as hostile by default. - Hunt the IIS-exclusion / rogue-local-user / "Google Chrome Start" scheduled-task triad and Nacos-config exfil as concrete artifacts of this chain.
- For the ViewState surface, audit IIS
machineKeyconfidentiality and alert on the inverted 5xx-success signature (500 containingInvalidCastExceptionon__VIEWSTATEPOST = exploit success).
Related pages
- UAT-10147 SPECTRE / BadIIS / agentic-AI campaign
- SPECTRE / Specter
- AI-augmented adversary operations
Sources
- Cisco Talos: UAT-10147: Chinese-speaking adversary integrates agentic AI into post-compromise operations (Joey Chen, August 20, 2026)
- Cisco Talos: UAT-10147 deploys SPECTRE: A cross-platform implant with Linux rootkit and BYOVD capabilities (Joey Chen, August 20, 2026)