keras vulnerabilities
Multi-backend Keras
- Registry
- PyPI
- Latest
- 3.15.1
- License
- Apache License 2.0
20 known advisories
The latest version, 3.15.1, is not affected by any of them. Projects on older versions may be.
pip install 'keras>=3.15.1'Is your project on an affected keras?
Look after several Python apps? Indie watches up to 15 repos, private ones too, for $5 a month. See Indie
Security advisories for keras
Keras model loading is vulnerable to denial of service through HDF5 shape bombs
- Affected
- <3.15.0
- Fixed in
- 3.15.0
Keras: HDF5 links can disclose local file contents
- Affected
- <3.12.3 · >=3.13.0 <3.15.0
- Fixed in
- 3.12.3, 3.15.0
Keras: TorchModuleWrapper can deserialize unsafe PyTorch pickle data
- Affected
- <3.12.3 · >=3.13.0 <3.15.0
- Fixed in
- 3.12.3, 3.15.0
Keras: tar extraction permits symlink-based path traversal
- Affected
- <3.12.3 · >=3.13.0 <3.15.0
- Fixed in
- 3.12.3, 3.15.0
Keras: Lambda deserialization can bypass safe mode and execute code
- Affected
- <3.12.3 · >=3.13.0 <3.15.0
- Fixed in
- 3.12.3, 3.15.0
Keras: HDF5 virtual datasets can disclose local files
- Affected
- <3.12.3 · >=3.13.0 <3.15.0
- Fixed in
- 3.12.3, 3.15.0
Keras: DiskIOStore permits path traversal through crafted layer names
- Affected
- <3.12.3 · >=3.13.0 <3.15.0
- Fixed in
- 3.12.3, 3.15.0
Keras archive extraction utilities allow path traversal and arbitrary file writes
- Affected
- <3.14.0
- Fixed in
- 3.14.0
Keras vulnerable to DoS via Malicious .keras Model (HDF5 Shape Bomb Causes Petabyte Allocation in KerasFileEditor)
- Affected
- >=3.0.0 <3.12.1 · >=3.13.0 <3.13.2
- Fixed in
- 3.12.1, 3.13.2
Keras has an untrusted deserialization vulnerability
- Affected
- <3.13.2
- Fixed in
- 3.13.2
Keras has a Local File Disclosure via HDF5 External Storage During Keras Weight Loading
- Affected
- >=3.13.0 <3.13.2 · >=3.0.0 <3.12.1
- Fixed in
- 3.12.1, 3.13.2
Keras Directory Traversal Vulnerability
- Affected
- <3.12.0
- Fixed in
- 3.12.0
Keras is vulnerable to arbitrary local file loading and Server-Side Request Forgery
- Affected
- <3.12.0
- Fixed in
- 3.12.0
Keras framework vulnerable to deserialization of untrusted data
- Affected
- >=3.11.0 <3.11.3
- Fixed in
- 3.11.3
The Keras `Model.load_model` method **silently** ignores `safe_mode=True` and allows arbitrary code execution when a `.h5`/`.hdf5` file is loaded.
- Affected
- >=3.0.0 <3.11.3
- Fixed in
- 3.11.3
Keras is vulnerable to Deserialization of Untrusted Data
- Affected
- <3.11.0
- Fixed in
- 3.11.0
Keras vulnerable to CVE-2025-1550 bypass via reuse of internal functionality
- Affected
- >=3.0.0 <3.11.0
- Fixed in
- 3.11.0
Arbitrary Code Execution via Crafted Keras Config for Model Loading
- Affected
- >=3.0.0 <3.9.0
- Fixed in
- 3.9.0
keras Path Traversal vulnerability
- Affected
- <=3.7.0
- Fixed in
- No fix released
Keras code injection vulnerability
- Affected
- <2.13.1rc0
- Fixed in
- 2.13.1rc0
Is your keras version vulnerable?
- keras 3.15No known advisories in 3.15.1
- keras 3.147 advisories in 3.14.1
- keras 3.138 advisories in 3.13.2
- keras 3.123 advisories in 3.12.4
- keras 3.1113 advisories in 3.11.3
- keras 3.1016 advisories in 3.10.0
- keras 2.1513 advisories in 2.15.0
- keras 2.1413 advisories in 2.14.0
- keras 2.1313 advisories in 2.13.1
- keras 2.1214 advisories in 2.12.0
- keras 2.1114 advisories in 2.11.0
- keras 2.1014 advisories in 2.10.0
- keras 1.214 advisories in 1.2.2
- keras 1.114 advisories in 1.1.2
- keras 1.014 advisories in 1.0.8
Which of your client projects still run an old keras?
NPMscan reads the requirements.txt, poetry.lock, uv.lock and Pipfile.lock in every GitHub repo you pick and checks each installed version against these advisories, including the packages your packages pull in. Each finding comes with the command that fixes it.
Other PyPI packages
Advisory data from OSV.dev (opens OSV.dev in a new tab), refreshed daily. Package details from the PyPI registry.