RabbitMQ EUVD-2026-87148

Exploitation requires the RabbitMQ management plugin to be configured with management.oauth_enabled=true - this makes the OAuth bootstrap endpoint reachable before authentication. The Authorization-header path requires the attacker to cause a victim's browser to submit an attacker-controlled bearer token that is reflected unescaped into the bootstrap JavaScript; the alternative cookie path additionally requires the attacker to first plant an access_token cookie on the management host (e.g., via a same-site sibling/subdomain or prior foothold). Active victim interaction (loading the affected management UI page) is required (UI:A). Limiting factors: non-default OAuth configuration, the extra cookie-planting prerequisite for that vector, and dependence on a victim admin session for meaningful impact. Affected versions are 4.2.0 through 4.2.7 and 4.3.0/4.3.1; fixed in 4.2.8 and 4.3.2.

This is a genuine but conditionally-gated reflected XSS (CWE-79) in the RabbitMQ management UI, and the raw 7.5 CVSS 4.0 score somewhat overstates the real-world priority. The vendor-supplied CVSS 4.0 vector (AV:N/AC:L/AT:P/PR:N/UI:A/VC:H/VI:H/VA:H) already encodes two dampeners the base score partly hides: AT:P (a specific attack requirement - the deployment must have management.oauth_enabled=true) and UI:A (active victim interaction is required). PR:N is technically correct because the OAuth bootstrap endpoint is reachable pre-authentication, but only in that OAuth-enabled configuration. No EPSS score, no CISA KEV listing, and no public POC were provided, so there is no evidence of active or automated exploitation - the fix is confirmed only by the vendor advisory and commits. The VC:H/VI:H/VA:H rating reflects that attacker JavaScript runs in the management UI origin (potential admin-session/credential theft and actions-as-victim), which is plausible for a broker admin console, though 'high' on all three impact dimensions is aggressive for what is fundamentally a client-side script-injection bug with no direct server confidentiality/integrity/availability compromise. Net: treat as a real bug to patch on OAuth-enabled management deployments, but not an internet-wide emergency - closer to a targeted, config-dependent issue than a mass-exploitation candidate.

An attacker crafts a request/link that causes a bearer token with embedded JavaScript to be reflected into the OAuth bootstrap script of a RabbitMQ management UI that has management.oauth_enabled=true; when an admin's browser processes it (Authorization header path), or after the attacker plants a malicious access_token cookie on the management host (cookie path), the unescaped token breaks out of the set_token_auth('...') call and executes in the management UI origin. No public exploit code was identified at time of analysis, though the fix commit clearly reveals the injection point.