CVE-2025-54468 Overview
CVE-2025-54468 is an information disclosure vulnerability in Rancher Manager. The flaw causes Rancher to forward Impersonate-Extra-* HTTP headers to external endpoints through the /meta/proxy endpoint. These headers can contain identifiable or sensitive user attributes such as email addresses and group memberships.
When Rancher proxies user-initiated requests to third-party services such as amazonaws.com, the impersonation headers are transmitted alongside the request. External parties receiving these requests can then collect user identity information that should remain internal to the Rancher cluster. The issue is tracked under CWE-200: Exposure of Sensitive Information to an Unauthorized Actor.
Critical Impact
User identity data — including email addresses and directory attributes — is leaked to external services proxied through Rancher Manager, enabling downstream tracking or reconnaissance.
Affected Products
- Rancher Manager (SUSE Rancher)
- Deployments exposing the /meta/proxy endpoint
- Multi-cluster Rancher environments proxying calls to external cloud APIs
Discovery Timeline
- 2025-10-02 - CVE-2025-54468 published to NVD
- 2026-06-17 - Last updated in NVD database
Technical Details for CVE-2025-54468
Vulnerability Analysis
Rancher Manager exposes a proxy endpoint at /meta/proxy that allows authenticated users to route requests to external services. Rancher supports Kubernetes user impersonation, which relies on HTTP headers such as Impersonate-User, Impersonate-Group, and Impersonate-Extra-* to carry the acting user's identity through the API server.
The vulnerability stems from Rancher failing to strip these impersonation headers before forwarding requests to non-Kubernetes destinations. When a user triggers a proxied call to a service such as amazonaws.com, the outbound request carries Impersonate-Extra-* headers containing user attributes. These attributes commonly include email addresses, principal IDs, and group claims sourced from the configured authentication provider.
Exploitation requires user interaction, and the confidentiality impact is limited to information the impersonation subsystem carries. However, the scope changes because leaked data crosses trust boundaries into third-party infrastructure operated outside the Rancher administrator's control.
Root Cause
The root cause is missing header sanitization in the /meta/proxy request handler. Rancher does not distinguish between internal Kubernetes API traffic — where impersonation headers are required — and external HTTP calls, where they must be stripped. This is a design flaw in how the proxy layer constructs outbound requests.
Attack Vector
An attacker capable of influencing a legitimate Rancher user to trigger a /meta/proxy request toward an attacker-controlled or attacker-monitored endpoint can passively collect the leaked headers. The attacker does not need credentials on the Rancher instance itself. Cloud providers or SaaS APIs receiving proxied requests may also log the headers, exposing identity data to any party with access to those logs.
For complete technical details, consult the GitHub Security Advisory GHSA-mjcp-rj3c-36fr and the SUSE Bugzilla entry for CVE-2025-54468.
Detection Methods for CVE-2025-54468
Indicators of Compromise
- Outbound requests from Rancher Manager to non-Kubernetes destinations containing Impersonate-Extra-*, Impersonate-User, or Impersonate-Group headers.
- /meta/proxy access log entries targeting external hostnames such as *.amazonaws.com or other public SaaS endpoints.
- Third-party service logs recording Kubernetes-style impersonation headers on inbound requests.
Detection Strategies
- Enable HTTP request logging on egress proxies or service meshes fronting Rancher and alert on Impersonate-* headers destined for external IP ranges.
- Review Rancher audit logs for /meta/proxy calls and correlate the target URL with cluster-internal versus external destinations.
- Inspect cloud provider API request logs (for example AWS CloudTrail request headers, if captured) for unexpected impersonation metadata.
Monitoring Recommendations
- Baseline the set of external hostnames Rancher legitimately proxies to and alert on deviations.
- Track user activity that generates high volumes of /meta/proxy calls, especially to newly observed destinations.
- Feed Rancher and egress proxy logs into a centralized analytics platform to enable correlation between user identity and outbound proxied traffic.
How to Mitigate CVE-2025-54468
Immediate Actions Required
- Upgrade Rancher Manager to a fixed release as identified in GHSA-mjcp-rj3c-36fr.
- Restrict which external destinations are reachable through /meta/proxy using an allowlist at the network egress layer.
- Audit historical /meta/proxy usage to identify which external services have already received impersonation headers.
Patch Information
SUSE and the Rancher project have issued fixes tracked in the vendor advisories. Refer to the GitHub Security Advisory GHSA-mjcp-rj3c-36fr for the exact fixed versions and to the SUSE Bugzilla entry for distribution-specific patch information. Apply upgrades in test environments before rolling to production clusters.
Workarounds
- Place an egress proxy in front of Rancher Manager that strips Impersonate-* headers from any request not destined for the Kubernetes API server.
- Block or restrict access to the /meta/proxy endpoint via ingress rules where the feature is not required.
- Rotate any user attributes — such as email aliases used as principal IDs — that may have been exposed to third parties through prior proxied requests.
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
