
CVE-2026-10520, CVE-2026-42271: PoeLLM Botnet — AI Server Cryptomining Surge (October 2026)
CVE-2026-10520, CVE-2026-42271 — PoeLLM Cryptomining Botnet
The high-severity vulnerabilities CVE-2026-10520 and CVE-2026-42271 are being actively exploited by the PoeLLM botnet to compromise over 3,400 AI and development servers. These flaws allow remote attackers to execute arbitrary code, deploy cryptomining payloads, and maintain persistent C2 communications. There is no official CVSS score yet, but exploitation is ongoing and widespread.
Attack Vector
PoeLLM targets servers running vulnerable AI/LLM services and open-source development tools. Attackers exploit these CVEs to gain initial access, then download a secondary payload that connects to command-and-control (C2) addresses hidden in a poem file (dash.css) within the GitHub repository ejejejdfbbebe. Infected hosts dynamically resolve new C2 infrastructure, enabling rapid botnet adaptation. The malware uses IPs such as 92.119.165.74 and 5.78.73.122 for C2 and lateral movement. Once established, the botnet mines cryptocurrency, scans for additional victims, and may attempt brute force attacks.
Who Is at Risk
At risk are organizations operating AI/LLM services and open-source development environments, particularly those exposed to the internet without recent patches. Most compromised servers are in North America and Western Europe. Any unpatched AI infrastructure or development server is a potential target, especially if running outdated or misconfigured open-source tools.
Patch & Mitigate
- Patch: Apply vendor patches for CVE-2026-10520 and CVE-2026-42271 immediately. Check for updates to all AI/LLM and related open-source components.
- Workaround: Restrict public access to AI and development servers. Disable unused services and enforce strong authentication.
- Detect: Monitor for outbound connections to 92.119.165.74, 5.78.73.122, and unusual GitHub traffic (repo ejejejdfbbebe, dash.css). Look for spikes in CPU usage and unauthorized process creation.
MITRE ATT&CK
- TA0001 — Initial Access: Attackers exploit exposed AI/LLM services to gain entry.
- TA0007 — Discovery: The botnet scans for additional vulnerable systems within the network.
- TA0009 — Collection: Infected hosts are used for cryptomining and further exploitation.
Source: https://www.helpnetsecurity.com/2026/10/08/poellm-malware-github-poem-ai-servers/
Start Your 14-Day Free Trial
Get curated cyber intelligence delivered to your inbox every morning at 6 AM. No credit card required.
Get Started Free

