LiteLLM, a proxy server for calling LLM APIs, prior to version 1.83.0, when configured with JWT authentication enabled, uses the first 20 characters of a JWT token as the cache key for OIDC userinfo. Because JWT headers generated by the same signing algorithm share identical initial characters, an attacker can create a token that matches the first 20 characters of a legitimate user's token, causing a cache hit and allowing the attacker to assume the legitimate user's identity and permissions. This improper authentication vulnerability (CWE-287) is not enabled by default and affects only those deployments with JWT/OIDC authentication enabled. The vulnerability has a CVSS 4.0 base score of 9.4 (critical). The issue is fixed in litellm version 1.83.0.

An unauthenticated attacker can impersonate a legitimate user by crafting a JWT token that matches the first 20 characters of a legitimate user's cached token, thereby inheriting that user's identity and permissions. This could lead to unauthorized access and privilege escalation in affected deployments with JWT/OIDC authentication enabled. The vulnerability does not affect instances where JWT authentication is not enabled.

A fixed version, litellm 1.83.0, is available and should be applied to remediate this vulnerability. Deployments that do not have JWT/OIDC authentication enabled are not affected. Patch status is confirmed by the vendor advisory. Users should upgrade to version 1.83.0 or later to resolve this issue.