/

CVE-2026-25767

Severity:

HIGH RISK

(8.1)

Improper Authorization in LavinMQ Policymaker Role

CVE-2026-25767

Analyzed

Vulnerability type

Category:

Improper Authorization

CWE:

CWE-863

Impact:

Cross-vhost Data Access and Unauthorized Message Publishing

Severity:

HIGH RISK

(8.1)

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

A flaw in LavinMQ's authorization logic allows authenticated users with the "Policymaker" management tag to create shovels that bypass normal access controls.

Under certain configurations, this can let those users read from or publish to virtual hosts (vhosts) they are not otherwise authorized to access, undermining isolation between tenants or applications.

Affected context

The vulnerability appears in LavinMQ deployments that:

  • Use management roles/tags, specifically the "Policymaker" tag
  • Rely on vhosts for tenant or application-level isolation
  • Allow authenticated users with the Policymaker tag to manage policies and shovels

In these environments, the authorization checks for shovel creation and configuration may not fully enforce per-vhost permissions, allowing shovels to be set up across vhosts the user cannot normally access.

Exploit preconditions

  • Attack surface: LavinMQ management interface or API endpoints that handle shovel and policy management.
  • Required access: Valid, authenticated LavinMQ account with the "Policymaker" management tag.
  • Configuration requirements: Multiple vhosts configured, with some vhosts the attacker is not authorized to read from or write to, and shovels enabled/usable in the deployment.
  • Trigger point: Creation or configuration of a shovel that specifies a source and/or destination vhost outside the user's authorized set.
  • Why exploitation is possible: The authorization layer enforces vhost permissions for normal client operations, but does not consistently apply these checks when a Policymaker user defines shovel endpoints, allowing indirect access via the shovel mechanism.

Why this happens

This issue stems from improper authorization (CWE-863), where a privileged management capability is not fully constrained by the normal access control model.

While direct client operations are checked against vhost-specific permissions, the shovel creation logic treats the Policymaker role as sufficiently trusted and fails to re-validate whether the user is allowed to read from or write to the specified vhosts.

Conceptually, this reflects a broken access control pattern where an indirect operation (configuring infrastructure objects like shovels) can be used to bypass the stricter checks applied to direct messaging operations.

Potential impact

  • Unauthorized reading of messages from vhosts the attacker is not permitted to access.
  • Unauthorized publishing of messages into restricted vhosts, potentially injecting malicious or misleading data.
  • Breach of tenant isolation in multi-tenant deployments, leading to cross-tenant data exposure.
  • Integrity impact on message-driven workflows, such as triggering unintended actions or corrupting message streams.
  • Expansion of an attacker's effective privilege within the messaging infrastructure via indirect routing paths.

General mitigation guidance

Mitigation typically focuses on tightening authorization rules around management capabilities and removing implicit trust in privileged tags.

Recommended defensive patterns include:

  • Ensuring that any configuration of shovels, policies, or routing objects is subject to the same vhost-level permission checks as direct client operations.
  • Restricting assignment of the "Policymaker" or similar high-impact management roles to a minimal set of administrative accounts, following least-privilege principles.
  • Separating duties so that operational accounts managing shovels do not also represent end-user or application identities.
  • Applying role and permission models that explicitly bind management actions to scoped resources (e.g., per-vhost administrative roles) rather than broadly trusted global roles.
  • Periodically reviewing access control policies and management tags to verify they align with isolation and tenancy requirements.

How teams detect issues like this

Detection approaches can focus on both configuration review and behavioral monitoring:

  • Configuration and policy review: Periodically audit management roles/tags and verify that only appropriate accounts hold the Policymaker tag or equivalent high-privilege capabilities.
  • Access control validation: Use security or QA checks that attempt to configure shovels targeting unauthorized vhosts to confirm that permission boundaries are enforced.
  • Log and telemetry analysis: Monitor management API logs for shovel creation or modification events where the source or destination vhost differs from the user's typical authorized scope or usage pattern.
  • Static and policy analysis: Apply configuration-as-code checks or policy-as-code tools to ensure that management permissions are tightly scoped and that indirect routing mechanisms cannot bridge restricted vhosts without explicit approval.

Sources

SourceLink
github.comgithub.com/cloudamqp/lavinmq/commit/3a83e5894495b60c7c32a79c3dbc9bd9fa237d9a
github.comgithub.com/cloudamqp/lavinmq/commit/be03da31f3db1a2552f7094ff58e953ef50cdc82
github.comgithub.com/cloudamqp/lavinmq/pull/1670
github.comgithub.com/cloudamqp/lavinmq/pull/1687
github.comgithub.com/cloudamqp/lavinmq/security/advisories/GHSA-wh37-6vrr-r9wg

Frequently asked

questions

Yes, but only by users who can authenticate to the LavinMQ management interface or API and who hold the Policymaker management tag. It is not exploitable by unauthenticated, anonymous users.

The attacker needs a valid LavinMQ account with the Policymaker tag or an equivalent management-level capability that allows them to define or modify shovels. Regular client accounts without this tag cannot directly exploit this flaw.

The safest pattern is to enforce consistent authorization checks across all operations, including management actions such as shovel creation. Shovel endpoints should only be accepted if the initiating user has explicit permissions on both the source and destination vhosts.

Organizations can prevent similar issues by designing access control models where management operations are subject to the same resource-scoped permission checks as direct data operations, by applying least privilege to administrative tags, and by including authorization boundary tests in automated security and integration testing.

Find out what's already
exploitable in your codebase.