Database/AI/ML frameworks & serving
vLLM (weight loading): `hf_model_weights_iterator` uses `torch.load` without `weights_only`
CVSS 7.5CVE-2025-24357AI/ML frameworks & servingcurated
Impact
hf_model_weights_iterator uses torch.load without weights_only → RCE
Who can reach it
Customer-supplied Hub checkpoint
What to do
Upgrade; a serving engine loading pickle weights inherits every PyTorch pickle CVE
References
Related entries
- vLLM (ZeroMQ): DoS and data exposure over ZeroMQCVE-2025-30202 · vLLM (ZeroMQ)High
- vLLM (HTTP GET): Single HTTP GET crashes the serverCVE-2025-48956 · vLLM (HTTP GET)High
- PyTorch (`torch.linalg.lu`): DoS on slice operationCVE-2025-55551 · PyTorch (`torch.linalg.lu`)High
- PyTorch (KV/conv path buffer overflow): Buffer overflow when a model combines Conv2d + hardshrink + viewCVE-2025-55558 · PyTorch (KV/conv path buffer overflow)High
- vLLM (API key comparison): Timing attack recovers the API keyCVE-2025-59425 · vLLM (API key comparison)High
- SGLang (`/server_info`): Endpoint returns API keys and SSL keyfile pathsCVE-2026-15977 · SGLang (`/server_info`)High
This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.