/

CVE-2026-56675

Severity:

HIGH RISK

(8.3)

Auth Bypass via Loopback Trust (CWE-287) in 9Router

CVE-2026-56675

Deferred

Vulnerability type

Category:

Authentication Bypass

CWE:

CWE-287, CWE-290, CWE-306, CWE-441

Impact:

Unauthorized Access to Protected APIs and Credential Abuse

Severity:

HIGH RISK

(8.3)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

9Router prior to version 0.5.2 incorrectly treats loopback-originated traffic as trusted, allowing unauthenticated access to /v1/* APIs when requests appear to come from 127.0.0.1.

A remote attacker can exploit reverse proxy setups that forward public traffic via localhost to access model listing and proxy endpoints, potentially abusing configured upstream provider credentials.

Affected context

This vulnerability occurs in 9Router deployments where:

  • The dashboard guard logic trusts requests with a loopback (127.0.0.1) source as local.
  • /v1/* endpoints are exposed to a reverse proxy or front-end server.
  • The reverse proxy forwards external traffic to the backend using 127.0.0.1 as the target.

In such configurations, external clients can be misclassified as local and bypass API key checks for /v1 APIs, including model discovery and provider proxy routes.

Exploit preconditions

  • Attack surface: Publicly reachable HTTP endpoints that front 9Router's /v1/* APIs.
  • Required access: Network-level ability to send HTTP requests to the reverse proxy or frontend that forwards traffic to 127.0.0.1 on the host running 9Router.
  • Trigger point: Requests to /v1/* (e.g., /v1/models, provider proxy endpoints) that arrive via a same-host reverse proxy and are evaluated by src/dashboardGuard.js.
  • Why exploitation is possible: The dashboard guard implicitly trusts loopback-originated traffic as local and skips API key enforcement, without verifying the true client origin or enforcing consistent authentication for protected endpoints.

Why this happens

The underlying bug is an over-trust of the loopback interface as an authentication signal.

The dashboard guard logic in 9Router assumes that any request arriving from 127.0.0.1 is local and thus trusted, allowing it to bypass API key checks for /v1/* routes.

When a reverse proxy or frontend server forwards external traffic to 9Router using the loopback address, the application misattributes the request's trust level.

This maps to CWE-287 (Improper Authentication), CWE-290 (Authentication Bypass by Spoofing), CWE-306 (Missing Authentication for Critical Function), and CWE-441 (Untrusted Proxy), where protection relies on network location or proxy behavior rather than explicit, verifiable credentials.

Potential impact

  • Remote unauthenticated access to /v1/models and other informational endpoints.
  • Abuse of configured upstream provider credentials through /v1 proxy endpoints.
  • Unauthorized consumption of paid or rate-limited AI API quotas.
  • Potential data exposure via model or provider metadata returned by /v1 APIs.
  • Facilitation of further misconfiguration exploitation in connected systems.

General mitigation guidance

Effective mitigation focuses on removing network-location trust assumptions and enforcing explicit authentication for sensitive APIs.

Common defensive patterns include:

  • Treating all requests, including those from loopback or internal networks, as untrusted until authenticated.
  • Enforcing API key or token-based authentication consistently across all /v1/* routes, regardless of source IP or proxy path.
  • Implementing clear separation between administrative endpoints and public-facing API routes, with stricter auth on administrative or credentialed operations.
  • Designing proxy integrations so that trust is based on signed tokens or mutual authentication, not on originating IP or host headers.
  • Applying least-privilege principles to upstream provider credentials and scoping them to the minimal necessary operations.

How teams detect issues like this

Organizations can detect similar authentication bypass patterns by:

  • Reviewing routing and guard logic for any trust decisions based on IP address, loopback checks, or assumptions about "local" traffic.
  • Employing static analysis or linters to flag conditional authentication paths where certain sources skip credential verification.
  • Conducting code reviews focused on access control around high-value endpoints (e.g., /v1/* APIs using external provider credentials).
  • Enforcing security policies that require uniform authentication for sensitive operations, regardless of proxy or network topology.
  • Using application logging and anomaly detection to highlight unauthenticated access to endpoints that should normally require keys or tokens.

Sources

SourceLink
github.comgithub.com/decolua/9router/commit/da667836cc7584bea0edd893de1d590c9ea279dc
github.comgithub.com/decolua/9router/releases/tag/v0.5.2
github.comgithub.com/decolua/9router/security/advisories/GHSA-x5c9-v98j-722r

Frequently asked

questions

Yes. If 9Router is deployed behind a reverse proxy that forwards public traffic via 127.0.0.1, a remote attacker can send requests through that proxy and be treated as a trusted local client.

The attacker only needs network access to the public-facing endpoint that ultimately routes to 9Router on localhost; they do not need an API key or prior authentication under the vulnerable configuration.

Require explicit authentication (API keys or tokens) for all sensitive /v1/* endpoints, regardless of source IP, and avoid relying on loopback or internal network status as an indicator of trust.

Use a consistent, credential-based authentication model; perform security reviews of proxy and routing logic; avoid using client IP or perceived locality as an authorization signal; and enforce access-control policies that treat all incoming traffic as untrusted until verified.

Find out what's already
exploitable in your codebase.