litellm vulnerable to remote code execution based on using eval unsafely

Critical9.8CVE-2024-5751 · Published Jun 27, 2024 · updated Sep 10, 2026

BerriAI/litellm version v1.35.8 contains a vulnerability where an attacker can achieve remote code execution. The vulnerability exists in the `add_deployment` function, which decodes and decrypts environment variables from base64 and assigns them to `os.environ`. An attacker can exploit this by sending a malicious payload to the `/config/update` endpoint, which is then processed and executed by the server when the `get_secret` function is triggered. This requires the server to use Google KMS and a database to store a model.