Home

Description

The Keras Model.load_model method can be exploited to achieve arbitrary code execution, even with safe_mode=True. One can create a specially crafted .keras model archive that, when loaded via Model.load_model, will trigger arbitrary code to be executed. This is achieved by crafting a special config.json (a file within the .keras archive) that will invoke keras.config.enable_unsafe_deserialization() to disable safe mode. Once safe mode is disable, one can use the Lambda layer feature of keras, which allows arbitrary Python code in the form of pickled code. Both can appear in the same archive. Simply the keras.config.enable_unsafe_deserialization() needs to appear first in the archive and the Lambda with arbitrary code needs to be second.

PUBLISHED Reserved 2025-09-03 | Published 2025-09-19 | Updated 2025-09-20 | Assigner Google




HIGH: 8.6CVSS:4.0/AV:L/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H/AU:Y/R:A

Problem types

CWE-502 Deserialization of Untrusted Data

Product status

Default status
unaffected

3.0.0 (semver) before 3.11.0
affected

Credits

Gabriele Digregorio finder

References

github.com/keras-team/keras/pull/21429 patch

cve.org (CVE-2025-9906)

nvd.nist.gov (CVE-2025-9906)

Download JSON