En bref
- CVE-2026-24747 est une vulnérabilité ÉLEVÉE (CVSS 8.8) touchant pytorch, torch, pytorch (linuxfoundation).
- Probabilité d'exploitation dans les 30 prochains jours (EPSS) : 0.80%, plus élevée que 56.5% des vulnérabilités.
- Correctif : mettre à jour vers 2.10.0 ou plus récent.
| Package | Affected versions | Patched versions |
|---|
| pytorch PyPI | < 2.10.0
| 2.10.0
|
| torch PyPI | < 2.10.0
| 2.10.0
|
| pytorch | < 2.10.0
| 2.10.0
|
Description
Summary
A vulnerability in PyTorch's weights_only unpickler allows an attacker to craft a malicious checkpoint file (.pth) that, when loaded with torch.load(..., weights_only=True), can corrupt memory and potentially lead to arbitrary code execution.
Vulnerability Details
The weights_only=True unpickler failed to properly validate pickle opcodes and storage metadata, allowing:
- Heap memory corruption via
SETITEM/SETITEMS opcodes applied to non-dictionary types
- Storage size mismatch between declared element count and actual data in the archive
Impact
An attacker who can convince a user to load a malicious checkpoint file may achieve arbitrary code execution in the context of the victim's process.
Credit
Ji'an Zhou
NVD description
PyTorch is a Python package that provides tensor computation. Prior to version 2.10.0, a vulnerability in PyTorch's `weights_only` unpickler allows an attacker to craft a malicious checkpoint file (`.pth`) that, when loaded with `torch.load(..., weights_only=True)`, can corrupt memory and potentially lead to arbitrary code execution. Version 2.10.0 fixes the issue.
Running one of these products? Scan your site to see whether it exposes a vulnerable version.
Scan my site free