# RabbitMQ Access Control Flaws Expose OAuth Secrets and Breach Tenant Isolation
## The Threat
RabbitMQ, one of the most widely deployed message brokers in enterprise infrastructure, contains two significant access control vulnerabilities that could allow attackers to leak OAuth client secrets and circumvent multi-tenant isolation boundaries. The flaws were discovered by Miggo's security team and disclosed on July 14, 2026, revealing a six-month window during which these issues have remained unpatched across production deployments globally.
The more critical vulnerability, tracked as CVE-2026-57219, stems from an obsolete HTTP API endpoint that exposes the broker's OAuth client secret in a single unauthenticated request. According to Miggo's analysis, attackers who obtain this secret can exchange it for an administrator token, granting full control over every message, queue, user, and broker configuration. This represents a direct path to complete broker compromise for any RabbitMQ installation configured with OAuth 2.0 using the management.oauth_client_secret key—a common pattern in cloud-deployed and multi-tenant environments.
The second vulnerability, CVE-2026-57221, affects authorization enforcement on sensitive management endpoints. The flaw allows any authenticated user with access to a virtual host to enumerate all queue and exchange names and read message counts and consumer statistics, regardless of their assigned permissions. Miggo's researchers noted that "the endpoint's authorization check was hard-coded to always allow the request, unlike every other sensitive management endpoint," indicating this was not an oversight but a fundamental design gap. Together, these flaws expose enterprises to cross-tenant data exposure, infrastructure reconnaissance, and broker takeover—risks that are most acute in cloud deployments and multi-tenant SaaS environments where tenant isolation is a core security requirement.
Both vulnerabilities have been present in RabbitMQ since early 2024, affecting all release lines from version 3.13.0 onward. There is currently no public evidence of active exploitation, but the relative simplicity of leveraging CVE-2026-57219 means attackers may already have begun probing for vulnerable instances on internet-accessible management interfaces.
## Severity and Impact
| Aspect | CVE-2026-57219 | CVE-2026-57221 |
|--------|---|---|
| CVE ID | CVE-2026-57219 | CVE-2026-57221 |
| CVSS Score | 8.7 (High) | 5.3 (Medium) |
| CVSS Vector | CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H | CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:N/A:N |
| Attack Vector | Network | Network |
| Attack Complexity | Low | Low |
| Authentication Required | None | Yes (low-privilege user) |
| Impact | Confidentiality, Integrity, Availability | Confidentiality (cross-tenant) |
| CWE | CWE-287 (Improper Authentication), CWE-640 (Weak Password Recovery) | CWE-639 (Authorization Bypass Through User-Controlled Key) |
## Affected Products
RabbitMQ versions impacted:
Fixed versions:
All release lines from version 3.13.0 onward are affected. Earlier versions (3.12 and prior) are not vulnerable.
## Mitigations
Immediate Actions:
1. Upgrade immediately to patched versions (3.13.15, 4.0.20, 4.1.11, 4.2.6, or 4.3.0) if you are running RabbitMQ 3.13.0 or later.
2. Rotate OAuth credentials if your RabbitMQ installation is configured with OAuth 2.0 and uses the management.oauth_client_secret configuration key. Assume the secret has been compromised if the management interface (port 15672) is or has been reachable from untrusted networks.
3. Restrict management port access by implementing network-level controls. Limit TCP port 15672 (RabbitMQ management UI) to administrative networks only. If the management interface must be exposed, place it behind a reverse proxy with strong authentication and encryption.
4. Implement firewall rules to block access to the vulnerable HTTP API endpoint (GET /api/auth) on unpatched instances if internal patching is delayed.
Architectural Mitigations:
5. Isolate tenants by virtual host with strict permission boundaries. Do not allow cross-tenant user accounts or shared credentials. Implement namespace-level isolation where possible.
6. Review user permissions on all virtual hosts. Remove overly broad permissions granted to service accounts or application users, particularly those with access to sensitive queue names or message counts.
7. Monitor management interface logs for unusual GET requests to /api/auth endpoint, which could indicate exploitation attempts. Flag any unauthenticated access to management endpoints.
## References
---
## HackWire Analysis
The RabbitMQ flaws highlight a recurring pattern in infrastructure security: the worst vulnerabilities are often not in new code, but in legacy endpoints that outlive their intended use. CVE-2026-57219 is a textbook example—an "obsolete" HTTP API endpoint that should have been deprecated long ago was instead left in the codebase without authorization checks, creating a six-month window of exposure for any cloud or multi-tenant operator who followed standard OAuth configuration patterns.
What makes this particularly sharp is the attack surface. Unlike vulnerabilities requiring sophisticated exploitation chains, obtaining an OAuth secret requires a single HTTP request to an unauthenticated endpoint. An attacker who discovers that a RabbitMQ management interface is internet-accessible—which happens frequently through misconfigured DNS, cloud metadata leakage, or deployment automation errors—can retrieve the secret in seconds and escalate to full broker control. The ease of exploitation is inversely proportional to the severity of the outcome, a combination that attracts active threat actors.
The second flaw, CVE-2026-57221, is equally telling. Authorization checks that are "hard-coded to always allow the request" suggest this endpoint was never designed with multi-tenant security in mind, or was retrofitted into a multi-tenant architecture without full security review. For organizations using RabbitMQ in hosted or SaaS deployments where tenant isolation is legally and operationally required, this creates a hidden compliance gap. Customers may believe their data is protected by virtual host boundaries, but a single authentication to any vhost grants visibility into the infrastructure of all tenants sharing that broker.
The timeline is also important: both flaws have been present since early 2024, meaning they predate the recent wave of critical RabbitMQ vulnerabilities (TLS bypass and JWT forgery flaws disclosed earlier this year). This suggests RabbitMQ's codebase has undergone recent security scrutiny that may now be exposing previously undetected issues. Organizations should expect additional disclosures and should prioritize upgrading sooner rather than waiting for a "safe" batch of patches.
For defenders, the immediate priority is access control: restrict management port exposure, rotate secrets, and audit permission assignments. For platform teams considering RabbitMQ, stronger defaults around authorization checks and mandatory deprecation of legacy endpoints should become non-negotiable acceptance criteria.
— *HackWire Editorial*
---
## Related Coverage