/

CVE-2026-45567

Severity:

HIGH RISK

(8.3)

Auth Bypass (CWE-287) in Roxy-WI API GPT Endpoint

CVE-2026-45567

Deferred

Vulnerability type

Category:

Authentication Bypass

CWE:

CWE-287, CWE-306, CWE-697

Impact:

Unauthorized Access to Protected Functionality

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

Roxy-WI versions 8.2.6.4 and earlier contain an authentication bypass in certain API URLs, notably the unauthenticated /api/gpt endpoint.

By abusing URL patterns containing the api substring, an attacker can reach functionality that should require prior login, creating a high-risk exposure on internet-facing Roxy-WI panels.

Affected context

This vulnerability affects Roxy-WI deployments up to and including version 8.2.6.4 where the web interface is reachable over HTTP(S).

It is most impactful when the Roxy-WI panel is exposed beyond strictly controlled administrative networks, or when network-layer controls assume the web UI enforces strong authentication on all management endpoints.

The issue arises in routes using api substring matching that incorrectly treat /api/gpt and similar paths as exempt from normal authentication checks.

Exploit preconditions

  • Attack surface: HTTP(S) interface of the Roxy-WI management UI.
  • Required access: Network-level reachability to the Roxy-WI web application; no prior authentication is required.
  • Trigger point: Direct HTTP requests to endpoints whose paths contain api, specifically the unauthenticated /api/gpt route.
  • Why exploitation is possible: Authentication logic is applied inconsistently, allowing some API endpoints to bypass standard authorization gates based on URL pattern handling rather than explicit, centralized access control checks.

Why this happens

The core issue maps to CWE-287: Improper Authentication, with related aspects of CWE-306: Missing Authentication for Critical Function and CWE-697: Incorrect Comparison.

The routing or middleware logic appears to rely on substring or pattern-based decisions (e.g., paths containing api) to treat certain endpoints differently, leading to missing or incorrect enforcement of login requirements for /api/gpt.

Instead of enforcing a uniform, centralized authentication and authorization check for all sensitive routes, the application uses conditional URL handling that unintentionally leaves some API functionality exposed to unauthenticated users.

Potential impact

  • Unauthenticated access to API functionality that may expose configuration or operational data.
  • Potential manipulation of management-related actions, depending on what /api/gpt and similar endpoints can perform.
  • Use of the exposed endpoint as a foothold for further reconnaissance against the Roxy-WI instance or underlying infrastructure.
  • Indirect impact on Haproxy, Nginx, Apache, or Keepalived configurations if the API allows interaction with those components.

General mitigation guidance

Effective mitigation typically focuses on enforcing consistent, centralized authentication and authorization for all management and API endpoints.

Recommended defensive patterns include:

  • Applying a single, mandatory authentication middleware for all routes under administrative or API namespaces, with explicit allowlists for truly public endpoints.
  • Avoiding substring-based or heuristic URL checks to determine whether authentication is required; instead, define access requirements declaratively per route or controller.
  • Implementing strict role- and permission-based checks for actions that can influence proxies, load balancers, or system configuration.
  • Restricting external exposure of the Roxy-WI interface using network controls (VPNs, bastion hosts, IP allowlists) and enforcing strong authentication factors at the perimeter.
  • Designing APIs so that potentially sensitive helpers (such as assistants or GPT-like endpoints) cannot access or manipulate privileged management state without explicit, verified user context.

How teams detect issues like this

Detection commonly involves a mix of application-layer review and external testing:

  • Static analysis / code review: Look for route handlers or middleware where authentication checks are skipped or conditioned on URL substrings, prefixes, or hostnames rather than explicit policy.
  • Configuration review: Enumerate all API routes (e.g., via framework routing tables) and verify that each is covered by a consistent authentication and authorization mechanism.
  • Dynamic testing: Use automated scanners or manual testing to probe API endpoints (such as /api/*) without credentials and confirm whether any respond with privileged data or actions.
  • Logging and monitoring: Enable detailed access logging for management endpoints and set alerts on unauthenticated requests to sensitive paths or anomalous patterns in API usage.

Sources

SourceLink
github.comgithub.com/roxy-wi/roxy-wi/security/advisories/GHSA-4fcm-qgg8-w2vf

Frequently asked

questions

Yes. If the Roxy-WI interface is reachable over the network, an attacker can exploit the unauthenticated API endpoints remotely without prior login.

No application-level credentials are required. The only prerequisite is network access to the affected Roxy-WI instance and knowledge or discovery of the vulnerable API paths.

The safest approach is to centralize authentication and authorization so that all management and API routes are protected by default, with explicit, minimal exceptions for endpoints that are intentionally public. Avoid URL substring logic for deciding whether authentication is necessary.

Adopt a defense-in-depth strategy: enforce secure-by-default route protection, define clear access control policies, add automated tests that verify all sensitive endpoints require authentication, and include security-focused code review rules to flag routes missing explicit authentication checks.

Find out what's already
exploitable in your codebase.