CRITICAL🇵🇱 Wersja polska

CVE-2026-48746

CVSS 9.1v3.1pub. 2026-06-22upd. 2026-08-14

vLLM is an inference and serving engine for large language models (LLMs). From 0.3.0 until 0.22.0, a vulnerability in ASGI web servers and starlette's trust on those web servers enables an authentication bypass of the OpenAI API AuthenticationMiddleware. It allows to use the API without providing the configured VLLM_API_KEY or --api-key. This vulnerability is fixed in 0.22.0.

🤖 AI Analysis
How it works

The vulnerability results from improper handling of HTTP requests by ASGI servers combined with incorrect trust by the Starlette framework of these servers (CWE-444: HTTP Request/Response Smuggling, CWE-501: Trust Boundary Violation). The AuthenticationMiddleware mechanism responsible for verifying the API key can be bypassed through a crafted network request. As a result, the middleware fails to enforce proper identity verification, allowing unauthorized access to protected API resources.

Impact

An attacker gains unauthorized access to the vLLM server API — can read data (C:H) and potentially cause service unavailability (A:H), without needing a valid API key.

Mitigation & patch

Update vLLM to version 0.22.0 or later. Red Hat errata are also available: RHSA-2026:30088 and RHSA-2026:30089. Until updating, it is recommended to restrict network access to vLLM API endpoints using a firewall or reverse proxy with its own authentication layer.

Who is affected

vLLM in versions 0.3.0 to 0.21.x (vulnerability fixed in version 0.22.0)

Analysis generated by Claude AI (Anthropic) based on NVD data. Always verify with vendor.
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:H
  • Vllm

    APP
    Vllm
    0.3.0 – 0.22.0 (excl.)
🔵
CHECK WITH VENDOR
No clear patch data available. Check vendor references.
Tags
Auth Bypass
CWE
References

Related vulnerabilities

CVE-2026-22778CRITICAL9.8PL ✓same product

vLLM: wyciek adresu sterty umożliwiający RCE przez endpoint multimodalny

CVE-2025-47277CRITICAL9.8PL ✓same product

vLLM: niezamierzone nasłuchiwanie TCPStore na wszystkich interfejsach sieciowych

CVE-2025-32444CRITICAL10.0PL ✓same product

RCE w vLLM poprzez deserializację pickle na niezabezpieczonych gniazdach ZeroMQ

CVE-2024-11041CRITICAL9.8PL ✓same product

RCE przez niebezpieczną deserializację w vllm MessageQueue.dequeue()

CVE-2025-29783CRITICAL9.0PL ✓same product

RCE przez niebezpieczną deserializację w vLLM z Mooncake (ZMQ/TCP)