CRITICALvulnerability·

Active Exploitation of LiteLLM RCE Vulnerabilities (CVE-2026-42271, CVE-2026-48710)

Threat actors are actively exploiting chained vulnerabilities, including CVE-2026-42271, in LiteLLM proxy servers to achieve unauthenticated remote code execution. This allows attackers to run arbitrary commands on affected AI gateway hosts, leading to potential cryptocurrency mining, botnet recruitment, or further compromise. Immediate patching is critical for vulnerable instances.

This report was researched and drafted by an AI agent and reviewed by a human analyst prior to publication. View the agent workflow →
Indicators of Compromise(37)
↓ Download STIX
Domainlitellm.proxy
Emailsecalert@redhat.com
IP92.119.164.50
IP103.249.201.108
IP178.128.14.204
IP191.37.28.160
IP89.39.253.46
IP120.224.114.212
IP5.78.73.122
IP15.204.178.28
IP92.119.165.74
IP45.133.73.28
IP185.132.53.158
IP136.148.69.233
SHA2560efd1bb73b277b4cdca47bcb424679de6fd094d04e73e7a3ed0c470e8040b440
SHA25676341c99cdb9a4321a37a58b1757ed0327bcf8d3adfa5688109f138a16b5ef04
SHA2561c962bba45e5cddaadd8ceff241417f9c3686aba1df8b6b511caf5a9901f2c40
SHA25672648350ddd7a3401542e90f923df38adb4ee75ded438dade1bc472566f474dc
SHA2565149bc20e5d0a9e281ca3330d000d613b5987f46996e771067532af0fe3e0fb8
SHA2564b8e9cda02c30b95aa6e483dd9f90b2e27b9093212c6c3d956be8da3dd00dd16
SHA2561d6d4da4451688faa350b05558ed601e5c498b2382923f1bf45fc41958b098f6
SHA256cc6f060d37c92ea68c955e637e6291863272c1992ba86bc872989c4b566fc8bf
SHA256305bf281a63bff13b96466fb7adfd0962ac28e92d649ad692a86eb10db89a1a0
SHA25666cc80e961fe991b063dc51467fa901dce3fb2afc383afed1c596ec432363022
SHA256c0d7ed557a77e880b7d78eb3a87a05cadaf8711eea4e024fece782f5a68edebe
SHA256a2362d90bb91fd360fdbfe760206da6a30024cca09b9839bdbfe0573a3e76ace
SHA25659a4a3dc01e258bdd6aa463c8dbb34e14bd1a8357ef6ae3d4ea9e6c8ce062353
SHA2561fc8714e90d4b7a9ca8e6d9360f692fee9a764d6e556d5241886dbf116df8902
SHA2565d74f4badc192e18e4cf8027a2091cf5148f1426e0542bc5d7a5ecaf6dd9cf87
SHA256227cecaf494ed3999aee9743b93371bd11eebce31538d34bea239133f473e3dd
SHA256b6ec1fac474c9ff596322446656add0ebb10c449d623c37a3f71548957dc4c9c
SHA256163f636a1cbc151572ad0470bd5b06650980efd72e6c462cff6ce9ce4bcfaa93
SHA2565b0afdcb5c35d4c55847e1eb2967056ef9f508549f40028d6a6d2b41f6e70fe6
SHA256ee821288511aaf6e91080f8a925a425e5d26eeacc73f042b39469c65c2e7a139
SHA256195a25181e1ff5c73cea1146a339489398cb54618e8cd9d574e90a72a1ff3e11
SHA25648a73c9e0491f2f9bb109e15fea89ebd65939ed9bcfced506e006f23bbe9f64e
SHA256f7788f465a7f7d03b8ac7b1de0994e99eb95693f78120856420e7d2426a3c466

Overview

Threat actors are actively scanning for and exploiting vulnerabilities in LiteLLM proxy servers, specifically CVE-2026-42271 and CVE-2026-48710. These chained vulnerabilities allow for unauthenticated remote code execution on affected AI gateway hosts. Successful exploitation can lead to arbitrary command execution, often leveraged for cryptocurrency mining or establishing botnet presence.

Technical Analysis

  • CVE-2026-42271 affects LiteLLM (AI Gateway) versions 1.74.2 through 1.83.6 (i.e., before 1.83.7). Red Hat OpenShift AI versions 2.25 (before 2.25.8), 3.3 (before 3.3.4), and 3.4 are also impacted.
  • The vulnerability lies in the POST /mcp-rest/test/connection and POST /mcp-rest/test/tools/list endpoints.
  • These endpoints accept a full server configuration in the request body, including command, args, and env fields used by the stdio transport.
  • When a stdio configuration is provided, the endpoints attempt to connect, which spawns the supplied command as a subprocess on the proxy host.
  • The subprocess executes with the privileges of the LiteLLM proxy process.
  • Authentication is required for CVE-2026-42271, but any valid proxy API key, including low-privilege internal-user keys, is sufficient. Horizon3.ai reports chaining with CVE-2026-48710 to achieve unauthenticated RCE.
  • The vulnerability is classified as CWE-77 (Improper Neutralization of Special Elements used in a Command) and CWE-78 (Improper Neutralization of Special Elements used in an OS Command).
  • CVSS 3.1 score: 8.8 (High), Vector: CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H.

Detection

  • Web Server Logs: Look for POST requests to /mcp-rest/test/connection or /mcp-rest/test/tools/list. Pay attention to the request body for unusual command, args, or env parameters.
  • Process Monitoring: Monitor for unusual child processes spawned by the litellm process or the web server process hosting LiteLLM. Examples include sh, bash, cmd.exe, powershell.exe, python, perl, wget, curl, nc, ncat, or known cryptocurrency miners (e.g., xmrig). For Linux systems, monitor auditd logs for similar process creation events.
  • Network Traffic: Monitor for outbound connections from the LiteLLM host to unusual IP addresses or domains, especially those associated with C2, cryptocurrency pools, or file downloads.
  • File System Monitoring: Look for suspicious file creations in web root directories or temporary directories, particularly executables, scripts, or known malware (e.g., PoeLLM, XMRig).

Sigma Detection Rules

⚠️ AI-generated detection rules. These are experimental starting points. Review field names, EventIDs, and logic against your environment’s schema before deploying. Tune to reduce false positives.

LiteLLM RCE Endpoint Access Attempt

title: LiteLLM RCE Endpoint Access Attempt
id: a1b2c3d4-e5f6-7890-1234-567890abcdef
status: experimental
description: Detects POST requests to known vulnerable LiteLLM endpoints, indicating potential exploitation attempts for CVE-2026-42271.
logsource:
  category: webserver
detection:
  selection:
    cs-method: 'POST'
    cs-uri-stem|contains:
      - '/mcp-rest/test/connection'
      - '/mcp-rest/test/tools/list'
  condition: selection
level: high

LiteLLM Suspicious Child Process Creation (Windows)

title: LiteLLM Suspicious Child Process Creation (Windows)
id: f0e9d8c7-b6a5-4321-fedc-ba9876543210
status: experimental
description: Detects suspicious process creation by the LiteLLM application on Windows, indicative of successful command execution via CVE-2026-42271.
logsource:
  product: windows
  service: sysmon
detection:
  selection_parent:
    ParentImage|endswith: '\litellm.exe'
  selection_child:
    Image|endswith:
      - '\cmd.exe'
      - '\powershell.exe'
      - '\pwsh.exe'
      - '\bash.exe'
      - '\sh.exe'
      - '\python.exe'
      - '\perl.exe'
      - '\wget.exe'
      - '\curl.exe'
      - '\nc.exe'
      - '\ncat.exe'
  condition: selection_parent and selection_child
level: critical

Mitigations

  1. Patch Immediately: Upgrade LiteLLM to version 1.83.7 or later. For Red Hat OpenShift AI users, apply relevant errata (e.g., RHSA-2026:27784, RHSA-2026:28960, RHSA-2026:30056).
  2. Network Segmentation: Isolate LiteLLM instances from critical internal networks to limit lateral movement in case of compromise.
  3. Least Privilege: Ensure the LiteLLM proxy process runs with the minimum necessary privileges to reduce the impact of successful command execution.
  4. API Key Management: Regularly rotate and audit API keys, especially internal-user keys, and restrict their access where possible.
  5. Input Validation: Implement robust input validation at the network edge, if possible, to filter out malicious command injection attempts in request bodies.

References

  • https://isc.sans.edu/diary/rss/33522
  • https://github.com/BerriAI/litellm/security/advisories/GHSA-v4p8-mg3p-g94g
  • https://github.com/BerriAI/litellm/releases/tag/v1.83.7-stable
  • https://access.redhat.com/security/cve/CVE-2026-42271
  • https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware
  • https://www.wiz.io/blog/ai-infrastructure-honeypot
  • https://ltna.com.au/cyber

Indicators of Compromise

Type Value Description
SHA256 0efd1bb73b277b4cdca47bcb424679de6fd094d04e73e7a3ed0c470e8040b440 SHA256 hash of odh-operator-bundle image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 163f636a1cbc151572ad0470bd5b06650980efd72e6c462cff6ce9ce4bcfaa93 Red Hat OpenShift AI image hash (odh-operator-bundle) (via https://access.redhat.com/errata/RHSA-2026:30056)
SHA256 195a25181e1ff5c73cea1146a339489398cb54618e8cd9d574e90a72a1ff3e11 Red Hat OpenShift AI image hash (odh-rhel9-operator) (via https://access.redhat.com/errata/RHSA-2026:30056)
SHA256 1c962bba45e5cddaadd8ceff241417f9c3686aba1df8b6b511caf5a9901f2c40 SHA256 hash of odh-trustyai-garak-lls-provider-dsp-rhel9 image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 1d6d4da4451688faa350b05558ed601e5c498b2382923f1bf45fc41958b098f6 SHA256 hash of odh-trustyai-garak-lls-provider-dsp-rhel9 image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 1fc8714e90d4b7a9ca8e6d9360f692fee9a764d6e556d5241886dbf116df8902 Hash for odh-rhel9-operator image (arm64) (via https://access.redhat.com/errata/RHSA-2026:28960)
SHA256 227cecaf494ed3999aee9743b93371bd11eebce31538d34bea239133f473e3dd Hash for odh-rhel9-operator image (s390x) (via https://access.redhat.com/errata/RHSA-2026:28960)
SHA256 305bf281a63bff13b96466fb7adfd0962ac28e92d649ad692a86eb10db89a1a0 SHA256 hash of odh-trustyai-garak-lls-provider-dsp-rhel9 image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 48a73c9e0491f2f9bb109e15fea89ebd65939ed9bcfced506e006f23bbe9f64e Red Hat OpenShift AI image hash (odh-rhel9-operator) (via https://access.redhat.com/errata/RHSA-2026:30056)
SHA256 4b8e9cda02c30b95aa6e483dd9f90b2e27b9093212c6c3d956be8da3dd00dd16 SHA256 hash of odh-rhel9-operator image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 5149bc20e5d0a9e281ca3330d000d613b5987f46996e771067532af0fe3e0fb8 SHA256 hash of odh-trustyai-garak-lls-provider-dsp-rhel9 image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 59a4a3dc01e258bdd6aa463c8dbb34e14bd1a8357ef6ae3d4ea9e6c8ce062353 Hash for odh-llama-stack-core-rhel9 image (arm64) (via https://access.redhat.com/errata/RHSA-2026:28960)
SHA256 5b0afdcb5c35d4c55847e1eb2967056ef9f508549f40028d6a6d2b41f6e70fe6 Red Hat OpenShift AI image hash (odh-rhel9-operator) (via https://access.redhat.com/errata/RHSA-2026:30056)
SHA256 5d74f4badc192e18e4cf8027a2091cf5148f1426e0542bc5d7a5ecaf6dd9cf87 Hash for odh-rhel9-operator image (ppc64le) (via https://access.redhat.com/errata/RHSA-2026:28960)
SHA256 66cc80e961fe991b063dc51467fa901dce3fb2afc383afed1c596ec432363022 Hash for odh-llama-stack-core-rhel9 image (amd64) (via https://access.redhat.com/errata/RHSA-2026:28960)
SHA256 72648350ddd7a3401542e90f923df38adb4ee75ded438dade1bc472566f474dc SHA256 hash of odh-rhel9-operator image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 76341c99cdb9a4321a37a58b1757ed0327bcf8d3adfa5688109f138a16b5ef04 SHA256 hash of odh-rhel9-operator image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 a2362d90bb91fd360fdbfe760206da6a30024cca09b9839bdbfe0573a3e76ace Hash for odh-rhel9-operator image (amd64) (via https://access.redhat.com/errata/RHSA-2026:28960)
SHA256 b6ec1fac474c9ff596322446656add0ebb10c449d623c37a3f71548957dc4c9c Red Hat OpenShift AI image hash (odh-llama-stack-core-rhel9) (via https://access.redhat.com/errata/RHSA-2026:30056)
SHA256 c0d7ed557a77e880b7d78eb3a87a05cadaf8711eea4e024fece782f5a68edebe Hash for odh-operator-bundle image (amd64) (via https://access.redhat.com/errata/RHSA-2026:28960)
SHA256 cc6f060d37c92ea68c955e637e6291863272c1992ba86bc872989c4b566fc8bf SHA256 hash of odh-rhel9-operator image (via https://access.redhat.com/errata/RHSA-2026:27784)
SHA256 ee821288511aaf6e91080f8a925a425e5d26eeacc73f042b39469c65c2e7a139 Red Hat OpenShift AI image hash (odh-llama-stack-core-rhel9) (via https://access.redhat.com/errata/RHSA-2026:30056)
SHA256 f7788f465a7f7d03b8ac7b1de0994e99eb95693f78120856420e7d2426a3c466 Red Hat OpenShift AI image hash (odh-rhel9-operator) (via https://access.redhat.com/errata/RHSA-2026:30056)
IP Address 103.249.201.108 Active PoeLLM C2 server (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 120.224.114.212 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 136.148.69.233 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 15.204.178.28 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 178.128.14.204 Active PoeLLM C2 server (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 185.132.53.158 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 191.37.28.160 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 45.133.73.28 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 5.78.73.122 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 89.39.253.46 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 92.119.164.50 Active PoeLLM C2 server (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
IP Address 92.119.165.74 PoeLLM campaign infrastructure (via https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware)
Domain litellm.proxy Observed ‘odd’ Host header in attack attempts targeting LiteLLM/Starlette vulnerabilities (via https://isc.sans.edu/diary/rss/33522)
Email secalert@redhat.com Red Hat security contact email (via https://access.redhat.com/errata/RHSA-2026:27784)

MITRE ATT&CK

🤖 AI Attribution
Generated by gemini-2.5-flash ·
2,574 input / 2,011 output tokens ·
Reviewed and approved by a human analyst before publication
#vulnerability#ai#botnet#critical#cryptocurrency-mining#cve#linux#litellm#rce#vulnerability#windows