Browse / Unsafe Model Loading and Supply Chain / CVE-2026-12484
Keras: TorchModuleWrapper can deserialize unsafe PyTorch pickle data
- Class
- Unsafe Model Loading and Supply Chain (filed by a model that read the advisory)
- Technique
- Unsafe deserialisation
- Affected
- keras (PyPI)
- Gets past
- Safe model loader or scanner
- CVSS
- CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
- EPSS
- 0.3% chance of exploitation in 30 days
How it works, in short
The TorchModuleWrapper config loader calls torch.load with unrestricted unpickling unless an ambient safe-mode flag is set, so untrusted configs run code. (Our summary, written by a model. The source text is below.)
Details
A vulnerability in keras-team/keras version 3.15.0 allows unsafe deserialization of attacker-controlled PyTorch pickle data through the public keras.layers.TorchModuleWrapper.from_config method. This method invokes torch.load(..., weights_only=False) without requiring an explicit unsafe opt-in, such as a safe_mode=False parameter. When called outside a SafeModeScope(True) context, the absence of an ambient safe mode state permits unsafe deserialization by default. This issue can lead to arbitrary code execution if untrusted Keras layer configurations are processed using this method. The vulnerability arises because the method does not enforce safe deserialization practices unless explicitly guarded by Keras safe mode.
References
- https://nvd.nist.gov/vuln/detail/CVE-2026-12484
- https://github.com/keras-team/keras/pull/23048
- https://github.com/keras-team/keras/pull/23165
- https://github.com/keras-team/keras/commit/55888d3becbdbb45dc16a55b489900f911e2dde5
- https://github.com/keras-team/keras/commit/d338a45204bdc787c8b3c4a9b82c1911cd52dedf
- https://github.com/keras-team/keras
- https://github.com/keras-team/keras/releases/tag/v3.12.3
- https://github.com/keras-team/keras/releases/tag/v3.15.0
- https://huntr.com/bounties/ab14df49-13b5-4442-b754-3189430bfa28
Source: GitHub Advisory Database, used under CC BY 4.0.