high CVSS 7.3

CVE-2025-9906·Python vulnerability

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.

Severity
high
Software
Python
Fixed in
3.11.0
Published
2025-09-19

Affected versions

From: 3.0.0

Until: 3.11.0

Fixed in: 3.11.0

How to fix this CVE

Update Python's Keras library to version 3.11.0 or later to patch the arbitrary code execution vulnerability in Model.load_model. Applications using Keras 3.0.0 through 3.11.0 should immediately upgrade, as attackers can bypass safe_mode protections by embedding malicious serialized code within .keras model archives. Verify the upgrade by checking the installed Keras version and testing model loading with untrusted files in an isolated environment.

sudo dnf update python3-keras

Defensia detects this vulnerability

How to check if you are affected

  1. Check installed Keras version: python3 -c "import keras; print(keras.__version__)"
  2. Identify services loading .keras model files: grep -r "Model.load_model\|keras.models.load_model" /opt /srv /home --include="*.py" 2>/dev/null
  3. Search application logs for model loading errors or unexpected deserialization calls: grep -i "load_model\|enable_unsafe_deserialization" /var/log/application.log
  4. Verify patch installation: python3 -c "import keras; exit(0 if keras.__version__ >= '3.11.0' else 1)" && echo 'Patched' || echo 'Vulnerable'

FAQ

What is CVE-2025-9906?

CVE-2025-9906 is an arbitrary code execution vulnerability in Keras Model.load_model that allows attackers to execute malicious Python code by embedding special configuration and pickled Lambda layer payloads within .keras archive files, even when safe_mode is enabled. The exploit works by triggering keras.config.enable_unsafe_deserialization() through a crafted config.json file.

Is CVE-2025-9906 being actively exploited?

No, CVE-2025-9906 is not currently listed as actively exploited in the CISA KEV catalog, and no public exploits are publicly available.

What versions of Python/Keras are affected by CVE-2025-9906?

Keras versions 3.0.0 through 3.11.0 are vulnerable. The vulnerability is fixed in Keras 3.11.0 and later. Python itself is not the vulnerable component; Keras is the affected library.

How do I check if my server is vulnerable to CVE-2025-9906?

Run: python3 -c "import keras; v = tuple(map(int, keras.__version__.split('.')[:2])); exit(0 if v < (3, 0) or v >= (3, 11) else 1)" — a return code of 1 indicates vulnerability.

Does Defensia detect CVE-2025-9906?

Yes — Defensia's CVE advisory scanner compares installed package versions against the NVD database. If Keras is installed on a monitored server, CVE-2025-9906 will appear in your dashboard with remediation steps.

Related Python CVEs

CVE-2026-25632CVSS 10EPyT-Flow is a Python package designed for the easy generation of hydraulic and water quality scenario data of water distribution networks. Prior to 0.16.1, EPyT-Flow’s REST API parses attacker-controlled JSON request bodies using a custom deserializer (my_load_from_json) that supports a type field. When type is present, the deserializer dynamically imports an attacker-specified module/class and instantiates it with attacker-supplied arguments. This allows invoking dangerous classes such as subprocess.Popen, which can lead to OS command execution during JSON parsing. This also affects the loading of JSON files. This vulnerability is fixed in 0.16.1.
CVE-2026-34938CVSS 10PraisonAI is a multi-agent teams system. Prior to version 1.5.90, execute_code() in praisonai-agents runs attacker-controlled Python inside a three-layer sandbox that can be fully bypassed by passing a str subclass with an overridden startswith() method to the _safe_getattr wrapper, achieving arbitrary OS command execution on the host. This issue has been patched in version 1.5.90.
CVE-2026-33054CVSS 10Mesop is a Python-based UI framework that allows users to build web applications. Versions 1.2.2 and below contain a Path Traversal vulnerability that allows any user supplying an untrusted state_token through the UI stream payload to arbitrarily target files on the disk under the standard file-based runtime backend. This can result in application denial of service (via crash loops when reading non-msgpack target files as configurations), or arbitrary file manipulation. This vulnerability heavily exposes systems hosted utilizing FileStateSessionBackend. Unauthorized malicious actors could interact with arbitrary payloads overwriting or explicitly removing underlying service resources natively outside the application bounds. This issue has been fixed in version 1.2.3.
CVE-2026-28505CVSS 10Tautulli is a Python based monitoring and tracking tool for Plex Media Server. Prior to version 2.17.0, the str_eval() function in notification_handler.py implements a sandboxed eval() for notification text templates. The sandbox attempts to restrict callable names by inspecting code.co_names of the compiled code object. However, co_names only contains names from the outer code object. When a lambda expression is used, it creates a nested code object whose attribute accesses are stored in code.co_consts, NOT in code.co_names. The sandbox never inspects nested code objects. This issue has been patched in version 2.17.0.
CVE-2026-26030CVSS 9.9Semantic Kernel, Microsoft's semantic kernel Python SDK, has a remote code execution vulnerability in versions prior to 1.39.4, specifically within the `InMemoryVectorStore` filter functionality. The problem has been fixed in version `python-1.39.4`. Users should upgrade this version or higher. As a workaround, avoid using `InMemoryVectorStore` for production scenarios.

References

Track CVEs across your fleet automatically

Defensia scans your Linux servers and tells you exactly which ones are running vulnerable versions — including CVE-2025-9906. Free for 1 server.

Get started free