/

CVE-2026-55196

Severity:

CRITICAL RISK

(9.1)

Auth Bypass (CWE-306) in Hermes WebUI Passkey Setup

CVE-2026-55196

Deferred

Vulnerability type

Category:

Authentication Bypass

CWE:

CWE-306

Impact:

Unauthorized Administrative Access

Severity:

CRITICAL RISK

(9.1)

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

Hermes WebUI contains an authentication bypass in its passkey registration flow that lets unauthenticated remote users register the first passkey credential.

When passkey login is enabled with no existing credentials, attackers can claim the initial passkey and gain lasting administrative control over the interface.

Affected context

The vulnerability arises in Hermes WebUI deployments that enable passkey-based authentication (HERMES_WEBUI_PASSKEY=1) and have no pre-existing passkey credentials configured.

In this initial state, specific passkey registration endpoints are reachable without login protection.

Environments where the WebUI is exposed to untrusted networks or the public internet are especially exposed, but internal deployments are also at risk if an attacker can reach the WebUI during its first-passkey setup window.

Exploit preconditions

  • Attack surface: HTTP(S) interface of Hermes WebUI
  • Required access: Network access to the WebUI when passkeys are enabled and no passkey credentials exist
  • Trigger point: Calling:
  • POST /api/auth/passkey/register/options
  • POST /api/auth/passkey/register

while the instance is in its initial, unclaimed passkey state

  • Why exploitation is possible: The passkey registration endpoints do not enforce authentication or an explicit enrollment gate when no credentials are present, allowing the first caller to register a passkey that is then treated as an administrative identity.

Why this happens

This issue is an authentication enforcement failure consistent with CWE-306: Missing Authentication for Critical Function.

The passkey registration endpoints perform a critical security function—binding the first passkey credential to the instance—but are exposed without requiring any prior authentication or shared secret when no credentials exist.

The design treats the "first registration" as a trusted bootstrap operation but does not apply any additional checks (such as one-time setup tokens, local-only binding, or admin pre-authorization), effectively allowing any reachable client to self-assign administrative access.

Potential impact

  • Remote attackers may register an initial passkey and become the de facto administrator.
  • Attackers could maintain persistent access by configuring additional accounts or credentials.
  • Configuration, data, and connected services managed via Hermes WebUI can be altered or disrupted.
  • The compromised WebUI may be leveraged for further lateral movement in the surrounding environment.

General mitigation guidance

Mitigation generally focuses on enforcing strong authentication and hardening the initial enrollment flow:

  • Require some form of authenticated or out-of-band verification (setup token, pre-shared secret, or existing admin login) before allowing passkey registration, even for the first credential.
  • Restrict access to administrative endpoints (including setup and registration) behind additional controls such as network allowlists, reverse proxies with separate authentication, or VPN-only exposure.
  • Treat bootstrap/setup modes as sensitive operations and design them to fail closed when environment variables are misconfigured or no admin identity is present.
  • Apply least privilege and separate duties so that compromise of a single credential does not automatically grant broad administrative power where possible.

How teams detect issues like this

Effective detection approaches typically include:

  • Static analysis and code review: Look for routes that perform security-sensitive operations (e.g., credential registration, configuration changes) without explicit authentication or authorization checks; flag endpoints that alter authentication state but lack guard middleware or access-control annotations.
  • Configuration and policy checks: Enforce policies requiring that any "first run" or bootstrap flows are protected by initialization tokens or limited to trusted networks; review environment configuration for features that implicitly enable unauthenticated setup.
  • Runtime monitoring and logging: Log all requests to authentication and enrollment endpoints with sufficient context (source IP, user agent, timestamps); alert on passkey registrations from unexpected sources or at unexpected times.
  • Automated security testing: Use API-focused security tests to probe for critical endpoints reachable without authentication, including during initial deployment states.

Sources

SourceLink
github.comgithub.com/nesquena/hermes-webui/commit/4d90577e25d5537cb07290eca3fb8abff3bab316
github.comgithub.com/nesquena/hermes-webui/pull/4171
github.comgithub.com/nesquena/hermes-webui/pull/4267
github.comgithub.com/nesquena/hermes-webui/releases/tag/v0.51.442
vulncheck.comvulncheck.com/advisories/hermes-webui-unauthenticated-passkey-registration-via-authentication-bypass

Frequently asked

questions

Yes. If the Hermes WebUI is network-accessible and passkey authentication is enabled with no existing credentials, any remote attacker who can reach the WebUI can attempt to claim the first passkey.

The attacker only needs network access to the affected WebUI instance during the unclaimed passkey state; no prior account, token, or authentication is required.

Require explicit authentication or a secure bootstrap mechanism before allowing passkey registration, and ensure that critical registration endpoints are always gated by robust access-control checks, even during first-time setup.

Treat setup and enrollment flows as high-risk operations: design them with secure defaults, require strong authentication or one-time initialization secrets, limit exposure to trusted networks, and review all authentication-related endpoints for missing or inconsistent access control during design and code review.

Find out what's already
exploitable in your codebase.