Disclosed weakness · no CVE recorded

LiteLLM Azure credential handling could expose secrets from the proxy

LiteLLM · LiteLLM Proxy with Azure OpenAI or Azure AI

First documented disclosure · Review checkpoint

Summary

An authenticated Azure model user could gain access to secrets held by the proxy.

Severity and exploitation evidence

High · 7.7/10 · CVSS 3.1. Severity source.

Exploitation not established: The original source establishes the weakness but does not report malicious exploitation.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The proxy accepted caller-supplied identity settings where the deployment administrator’s settings should have prevailed.

Conditions required

A low-privilege user can access an Azure model on a proxy without Azure API-key environment settings.

Potential impact

Exposed credentials could grant upstream access or proxy administration, depending on which secrets the deployment held.

Affected versions and configurations

Affected builds precede the branch-specific fixes named by the project.

Fix and mitigation status

1.106.9, 1.108.6, 1.109.1 or 1.110.0-rc.2 on the corresponding release line.

Practical next steps

Install the fix for the deployed branch. Until then, use the gateway policy described by LiteLLM; review and rotate secrets if exposure is suspected.

Operating lesson

Keep provider identity configuration under operator control even when the caller is authenticated.

Evidence and limits

The project published the request-field change on November 11 and the security advisory on November 12. No CVE was shown in the advisory.

Sources