Broken Access Control in lunary-ai/lunary

Vulnerability

Overview CVE-2024-6582 is a broken access control vulnerability in lunary-ai/lunary, an open-source AI tool. The flaw resides in the saml.ts file, where the authorization mechanism fails to properly verify that a user belongs to the organization whose Identity Provider (IDP) settings or SSO metadata they are modifying or viewing [1]. Lacking proper tenant scoping, any authenticated user can target an arbitrary organization's SAML configuration.

Exploitation

An attacker must be an authenticated user within the lunary application. No special privileges are required beyond being logged in. By crafting requests to the affected endpoint, an attacker can update the IDP settings—such as certificate or SSO URL—and read the SSO metadata of another organization if they know the target organization's identifier [1]. This can be performed without any interaction from the victim organization's users.

Impact

Successful exploitation allows the attacker to alter the SAML authentication flow for the target organization. By modifying IDP settings, the attacker can redirect users to a malicious identity provider or disable security checks. If combined with knowledge of a target user's email address, this can lead to account takeover because the attacker could authenticate as that user via the compromised SSO [1]. This undermines trust within multi-tenant deployments.

Mitigation

As of the publication date (2024-09-13) there is no confirmed patch for CVE-2024-6582. The vulnerability was reported via the Huntr bounty program, and users should monitor the lunary repository for patches or implement strict tenant isolation controls [2]. No workaround is currently documented.