Yuxi, a large-model-based knowledge base and graph agent platform, suffers from an improper authentication vulnerability (CWE-287) in versions before 0.6.2. The platform only performs a validity check on the identity token in the Authorization header without verifying its origin or uniqueness per deployment. Consequently, administrator tokens generated in other instances or local environments can be reused to access backend management interfaces on different affected instances. An attacker with such a token can bypass normal authentication, invoke backend APIs, create administrator accounts, and take over the system. This vulnerability has been addressed in version 0.6.2. Temporary mitigations include setting a unique, strong JWT_SECRET_KEY per deployment and restricting public exposure of backend interfaces.

An attacker who obtains or crafts an administrator Authorization token valid in another deployment can bypass authentication controls and gain administrator privileges on an affected Yuxi instance. This allows unauthorized access to system configurations, backend management APIs, and the ability to create administrator accounts, leading to full system backend compromise. The CVSS score of 9.4 reflects the critical impact with network attack vector, no privileges required, no user interaction, and high confidentiality, integrity, and low availability impacts.

A fix is available in Yuxi version 0.6.2. Users should upgrade to this version to fully remediate the vulnerability. Until upgrading, users are advised to set the environment variable JWT_SECRET_KEY to a unique, strong value per deployment instance to prevent token reuse across deployments. Additionally, avoid exposing backend management interfaces directly to the public network to reduce attack surface.