Browse / Unsafe Model Loading and Supply Chain / CVE-2025-7647
llama-index-core insecurely handles temporary files
- Class
- Unsafe Model Loading and Supply Chain (filed by a model that read the advisory)
- Technique
- Other weaknesses
- Also involves
- Data and Model Poisoning
- Affected
- llama-index-core (PyPI)
- CVSS
- CVSS:3.0/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:L
- EPSS
- 0.1% chance of exploitation in 30 days
How it works, in short
A predictable shared temp cache directory lets other local users read or replace cached models and embeddings, or plant symlinks. (Our summary, written by a model. The source text is below.)
Details
The llama-index-core package, up to version 0.12.44, contains a vulnerability in the get_cache_dir() function where a predictable, hardcoded directory path /tmp/llama_index is used on Linux systems without proper security controls. This vulnerability allows attackers on multi-user systems to steal proprietary models, poison cached embeddings, or conduct symlink attacks. The issue affects all Linux deployments where multiple users share the same system. The vulnerability is classified under CWE-379, CWE-377, and CWE-367, indicating insecure temporary file creation and potential race conditions.
References
- https://nvd.nist.gov/vuln/detail/CVE-2025-7647
- https://github.com/run-llama/llama_index/commit/98816394d57c7f53f847ed7b60725e69d0e7aae4
- https://github.com/run-llama/llama_index
- https://huntr.com/bounties/a2baa08f-98bf-47a8-ac83-06f7411afd9e
Source: GitHub Advisory Database, used under CC BY 4.0.