CVE-2025-24357
vllm: Malicious model to RCE by torch.load in hf_model_weights_iterator
7.5
HIGH
CVSS 3.1
EPSS 1.0%
Description
vLLM is a library for LLM inference and serving. vllm/model_executor/weight_utils.py implements hf_model_weights_iterator to load the model checkpoint, which is downloaded from huggingface. It uses the torch.load function and the weights_only parameter defaults to False. When torch.load loads malicious pickle data, it will execute arbitrary code during unpickling. This vulnerability is fixed in v0.7.0.
How to fix CVE-2025-24357
To remediate CVE-2025-24357, upgrade the affected package to a fixed version below.
- —upgrade to 0.7.0 or later
- —upgrade to d3d6bb13fb62da3234addf6574922a4ec0513d04 or later
Is CVE-2025-24357 being exploited?
Low — EPSS is 1.0%, meaning exploitation activity has not been observed at scale.
Affected packages (2)
- from 0, < 0.7.0
- from 0, < d3d6bb13fb62da3234addf6574922a4ec0513d04 | from 0, < 0.7.0
CVSS scores
| Source | Version | Severity | Vector |
|---|---|---|---|
| osv | CVSS 3.1 | HIGH7.5 | CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:H/I:H/A:H |