CVE-2026-12484: Keras: TorchModuleWrapper can deserialize unsafe PyTorch pickle data
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.fromconfig method. This method invokes torch.load(..., weightsonly=False) without requiring an explicit unsafe opt-in, such as a safemode=False parameter.
Published Jul 19, 2026 · Updated Aug 7, 2026
Affected packages
keras
- Ecosystem
- PyPI
- Affected
- < 3.12.3
- Fixed in
- 3.12.3
- Check
pip show keras
keras
- Ecosystem
- PyPI
- Affected
- >= 3.13.0, < 3.15.0
- Fixed in
- 3.15.0
- Check
pip show keras
Advisory details
From the GitHub Advisory Database (CC BY 4.0).
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://huntr.com/bounties/ab14df49-13b5-4442-b754-3189430bfa28
- 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/releases/tag/v3.12.3
- https://github.com/keras-team/keras/releases/tag/v3.15.0
- https://github.com/advisories/GHSA-v2w2-w228-c444
How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. No analysis has been written for this entry yet. Methodology
More PyPI vulnerabilities
All PyPI entriesCVE-2026-12570
Keras model loading is vulnerable to denial of service through HDF5 shape bombs
medium5.5Aug 10, 2026CVE-2026-9335
Keras: HDF5 links can disclose local file contents
medium6.5Aug 2, 2026CVE-2026-12482
Keras: tar extraction permits symlink-based path traversal
low3.1Jul 14, 2026CVE-2026-12481
Keras: Lambda deserialization can bypass safe mode and execute code
high8.8Jul 3, 2026CVE-2026-12480
Keras: HDF5 virtual datasets can disclose local files
medium5.5Jul 1, 2026CVE-2026-12479
Keras: DiskIOStore permits path traversal through crafted layer names
medium6.1Jun 22, 2026
