Browse / Unsafe Model Loading and Supply Chain / CVE-2024-2952
LiteLLM has Server-Side Template Injection vulnerability in /completions endpoint
- Class
- Unsafe Model Loading and Supply Chain (filed by a model that read the advisory)
- Technique
- Code and command execution
- Affected
- litellm (PyPI)
- CVSS
- CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
- EPSS
- 1.3% chance of exploitation in 30 days
How it works, in short
A malicious tokenizer_config.json supplies a chat_template that LiteLLM renders with Jinja unsafely, executing code on the server when the model is used. (Our summary, written by a model. The source text is below.)
Details
BerriAI/litellm is vulnerable to Server-Side Template Injection (SSTI) via the /completions endpoint. The vulnerability arises from the hf_chat_template method processing the chat_template parameter from the tokenizer_config.json file through the Jinja template engine without proper sanitization. Attackers can exploit this by crafting malicious tokenizer_config.json files that execute arbitrary code on the server.
References
- https://nvd.nist.gov/vuln/detail/CVE-2024-2952
- https://github.com/BerriAI/litellm/issues/2949
- https://github.com/BerriAI/litellm/pull/2941
- https://github.com/BerriAI/litellm/commit/8a1cdc901708b07b7ff4eca20f9cb0f1f0e8d0b3
- https://github.com/BerriAI/litellm
- https://github.com/BerriAI/litellm/blob/0d803e13798db40aa7463e64a6bafaee386424f5/litellm/proxy/proxy_server.py#L2087
- https://github.com/advisories/GHSA-46cm-pfwv-cgf8
- https://github.com/pypa/advisory-database/tree/main/vulns/litellm/PYSEC-2026-387.yaml
- https://huntr.com/bounties/a9e0a164-6de0-43a4-a640-0cbfb54220a4
- https://pypi.org/project/litellm
Source: GitHub Advisory Database, used under CC BY 4.0.