BerriAI has addressed a significant authentication vulnerability in LiteLLM that allowed unauthorized access to its MCP tooling, impacting user security.
What changed
BerriAI has released version 1.84.0 of LiteLLM, which fixes a serious improper authentication vulnerability. This flaw allowed unauthenticated attackers to exploit the MCP Streamable HTTP endpoint by using a manipulated Authorization header.
- The vulnerability replaced the necessary key validation with an empty UserAPIKeyAuth() object.
- This change permitted unauthorized requests to interact with MCP tooling, bypassing security measures.
Who is affected
The vulnerability affects all users of BerriAI LiteLLM prior to version 1.84.0, particularly those utilizing the MCP Streamable HTTP endpoint.
Organizations relying on LiteLLM for API calls to OpenAI or other native formats are at risk.
- BerriAI LiteLLM users who have not updated to the latest version.
- Developers integrating with OpenAI APIs through LiteLLM.
- Businesses that depend on secure API interactions for their applications.
What to do
It is crucial for users of BerriAI LiteLLM to upgrade to version 1.84.0 immediately to mitigate the risk posed by this vulnerability.
Additionally, organizations should review their security protocols and ensure that proper authentication measures are in place.
- Update LiteLLM to version 1.84.0 or later.
- Conduct a security audit of API access and authentication mechanisms.
- Educate teams about the importance of keeping software up to date to prevent similar vulnerabilities.