A vulnerability has been identified within Rancher Manager whereby the SAML authentication from the Rancher CLI tool is vulnerable to phishing attacks. The custom authentication protocol for SAML-based providers can be abused to steal Rancher’s authentication tokens.
Rancher Manager deployments without SAML authentication enabled are not affected by this vulnerability.
An attacker can generate a phishing SAML login URL which contains a publicKey
and requestId
controlled by the attacker. The attacker can then give the link to another user (eg: admin) and if the victim goes to the link unsuspectingly, they might not notice the bad parameters in the URL. The user will be prompted to login and might believe that its session has ended so they need to re-login. By clicking on the link, they will be logged in and an encrypted token will be created with the attacker's public key. The attacker can then decrypt the victim’s Rancher token, enabling the attack
Please consult the associated MITRE ATTACK Techniques - Privilege Escalation for further information about this category of attack.
This vulnerability is addressed by changes in both Rancher Manager and the CLI. The fixed versions:
requestId
that is relevant to the authentication session in process as a separate message to the terminal, making it easier for the user to identify and compare.requestId
, enabling users to verify whether the request was intended or not.Patched versions of Rancher include releases v2.12.2, v2.11.6, v2.10.10, and v2.9.12.
If you cannot update to one of the fixed versions, make sure to check the URL printed by the Rancher CLI when performing a SAML authentication flow, especially the requestId
parameter. The URL and the requestId
printed by the CLI must be the same URL that you access and see when finalizing the SAML authentication flow in your browser for the Rancher UI. The URL has the format:
<rancher_ui_address>/dashboard/auth/login?requestId=<requestid>&publicKey=<pubkey>&responseType=<responsetype>
If the URL presented in the browser is different from the one logged by the CLI, the user must not proceed with the login, as they might be under a phishing attack.
If you have any questions or comments about this advisory:
{ "cwe_ids": [ "CWE-345" ], "nvd_published_at": "2025-10-02T12:15:28Z", "severity": "HIGH", "github_reviewed": true, "github_reviewed_at": "2025-09-26T13:02:44Z" }