vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server
Impact
The OpenAI-compatible server registers Starlette CORSMiddleware with allowed_origins defaulting to [""], and performs no authentication unless --api-key (or VLLM_API_KEY) is configured. A web page on any origin can send requests to a vLLM server that the victim's browser can reach (for example localhost or an internal/VPN address) and read the responses. The default allow_credentials=False blocks cookie-bearing cross-origin requests; because vLLM authenticates with a bearer header rather than cookies, the wildcard still permits untokenized access to an unauthenticated server and read…
Source excerpt · OSV/v1/load_lora_adapter and
/v1/unload_lora_adapter are inside /v1 and therefore do require a key when
one is set; they are open only when no key is configured and dynamic LoRA
updating is enabled.
Affected packages
Start with the recorded fix for your release branch. Check your installed version and test compatibility before upgrading. Fixing this advisory does not rule out other vulnerabilities in the release.
| Affected range | Recorded fix |
|---|---|
| Earlier releases before 0.32.0Ecosystem versions | 0.32.0 |
Version evidence: OSV · GHSA-vvpg-rrww-wpmhOSV · PYSEC-2026-4348
View original version evidence (198 explicit versions)
Affected products
Vendor version evidence. Product records are separate from automatic dependency matching.
vllm-project · GitHub_M
vllm
1 range
vllm-project · GitHub_M
vllm
Compare with this publisher’s ranges. This does not inspect your installation.
affected
< 0.32.0
Other versions: unknown
Original source description
Summary
The OpenAI-compatible server registers Starlette CORSMiddleware with
allowed_origins defaulting to ["*"], and performs no authentication unless
--api-key (or VLLM_API_KEY) is configured. A web page on any origin can send
requests to a vLLM server that the victim's browser can reach (for example
localhost or an internal/VPN address) and read the responses. The default
allow_credentials=False blocks cookie-bearing cross-origin requests; because
vLLM authenticates with a bearer header rather than cookies, the wildcard still
permits untokenized access to an unauthenticated server and read access when the
caller already holds a token.
Details
Affected components
vllm/entrypoints/launchers/cli_args.py-allowed_originsdefault value.vllm/entrypoints/serve/middleware/register.py-CORSMiddlewareregistration.vllm/entrypoints/serve/middleware/authenticate.py- auth middleware and its guarded path prefixes.
Root cause
vllm/entrypoints/launchers/cli_args.py:
allow_credentials: bool = False
"""Allow credentials."""
allowed_origins: list[str] = field(default_factory=lambda: ["*"])
"""Allowed origins."""
vllm/entrypoints/serve/middleware/register.py:
app.add_middleware(
CORSMiddleware,
allow_origins=args.allowed_origins,
allow_credentials=args.allow_credentials,
allow_methods=args.allowed_methods,
allow_headers=args.allowed_headers,
)
The auth middleware is added only when a token is configured and guards only the
/v1, /v2, /inference, and /cohere prefixes. /v1/load_lora_adapter and
/v1/unload_lora_adapter are inside /v1 and therefore do require a key when
one is set; they are open only when no key is configured and dynamic LoRA
updating is enabled.
Impact
- A malicious page can enumerate models (
GET /v1/models), issuePOST /v1/chat/completions, consume GPU time, and read model output through the victim's browser. - When the server is reachable only from the victim's network, the browser acts as a proxy a remote website could not otherwise use.
- With
allow_credentials=False, cookie-based CSRF does not apply; the exposure is cross-origin request execution and response disclosure.
This advisory was fixed in MR61009.
Inside the record
Evidence available
Describes the evidence in this view, not whether the product is safe.
Original records
CVE-2026-115917View the latest unmodified CVE JSON from the official source. It may be newer than this imported page.
cve.new JSON →Underlying weaknesses
The source’s classification of what caused the vulnerability.
Contributing organizations · 2
Assigning organization
Additional data publisher
What changed
Follow changes- CVE · CVE-2026-115917
Observed 2026-10-31T01:57:41.336Z · upstream updated 2026-10-31T01:38:53.155Z
Prioritization updated
before
[]
after
[ { "source": "CISA-ADP", "type": "SSVC", "role": "CISA Coordinator", "timestamp": "2026-10-31T01:38:43.564283Z", "decisions": [ { "name": "Exploitation", "value": "none" }, { "name": "Automatable", "value": "no" }, { "name": "Technical Impact", "value": "partial" } ] } ]
Large field values are marked as truncated previews. Original source snapshots remain retained.
Showing up to 50 most recent observed changes.