CVE-2025-9906

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.

Weakness

The product deserializes untrusted data without sufficiently ensuring that the resulting data will be valid.

Affected Software

Potential Mitigations

• Make fields transient to protect them from deserialization.
• An attempt to serialize and then deserialize a class containing transient fields will result in NULLs where the transient data should be. This is an excellent way to prevent time, environment-based, or sensitive variables from being carried over and used improperly.

Related Attack Patterns

• https://cwe.mitre.org/data/definitions/586.html

References

• https://github.com/keras-team/keras/pull/21429