CVE-2026-12570
Denial of Service via HDF5 Shape Bomb in keras.models.load_model() in keras-team/keras
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
Attack profile
How easy it is to exploit
- Attack vector
- Local
- Complexity
- Low
- Privileges needed
- None
- User interaction
- Required
What an attacker can affect
- Confidentiality
- None
- Integrity
- None
- Availability
- High
Exploit likelihood
What each source reports
CVE List Score used
Denial of Service via HDF5 Shape Bomb in keras.models.load_model() in keras-team/keras
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
CVE-2026-12570- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Weakness
- CWE-770
- Updated
- 2026-11-02
EUVDEUVD-2026-55002
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
EUVD-2026-55002- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Updated
- 2026-11-02
GitHub AdvisoriesGHSA-74m6-m3xx-3vmj
Keras model loading is vulnerable to denial of service through HDF5 shape bombs
A vulnerability in keras-team/keras versions < 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
GHSA-74m6-m3xx-3vmj- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Weakness
- CWE-770
- Updated
- 2026-11-02
NVD
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
CVE-2026-12570- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Weakness
- CWE-770
- Updated
- 2026-11-02
OSV
Denial of Service via HDF5 Shape Bomb in keras.models.load_model() in keras-team/keras
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
CVE-2026-12570- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Weakness
- CWE-770
- Updated
- 2026-11-02
OSVDEBIAN-CVE-2026-12570
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
DEBIAN-CVE-2026-12570- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Updated
- 2026-11-02
Show 6 moreShow less
OSVGHSA-74m6-m3xx-3vmj
Keras model loading is vulnerable to denial of service through HDF5 shape bombs
A vulnerability in keras-team/keras versions < 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
GHSA-74m6-m3xx-3vmj- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Weakness
- CWE-770
- Updated
- 2026-11-02
OSVOESA-2026-3952
python-Keras security update
Keras is a high-level neural networks API for Python.
Security Fix(es):
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.(CVE-2026-12570)
- Record
OESA-2026-3952- CVSS 3.1
- 5.5 Medium
CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Updated
- 2026-11-02
OSVOESA-2026-3953
python-Keras security update
Keras is a high-level neural networks API for Python.
Security Fix(es):
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.(CVE-2026-12570)
- Record
OESA-2026-3953- CVSS 3.1
- 5.5 Medium
CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Updated
- 2026-11-02
OSVOESA-2026-4052
python-Keras security update
Keras is a high-level neural networks API for Python.
Security Fix(es):
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.(CVE-2026-12570)
- Record
OESA-2026-4052- CVSS 3.1
- 5.5 Medium
CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Updated
- 2026-11-02
OSVPYSEC-2026-3856
Keras model loading is vulnerable to denial of service through HDF5 shape bombs
A vulnerability in keras-team/keras versions < 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
PYSEC-2026-3856- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Updated
- 2026-11-02
OSVUBUNTU-CVE-2026-12570
A vulnerability in keras-team/keras versions <= 3.15.0 allows for a denial of service (DoS) attack when loading malicious .keras model files via the keras.models.load_model() function. The H5IOStore.getitem method in keras/src/saving/saving_lib.py does not validate the shape or size of datasets, leading to unbounded memory allocation. A specially crafted .keras file can exploit this flaw to trigger an out-of-memory (OOM) condition, causing the process to be terminated (exit code 137). This issue bypasses the fix for CVE-2026-0897, which only addressed a similar vulnerability in KerasFileEditor. The attack vector includes poisoned models from public repositories or malicious model registries, posing a risk to machine learning pipelines that process untrusted models.
- Record
UBUNTU-CVE-2026-12570- CVSS 3.0
- 5.5 Medium
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H - Updated
- 2026-11-02
Affected products
Affected packages
| Package | Affected | Fixed |
|---|---|---|
kerasPyPI |
before 3.15.0 | 3.15.0 |
kerasDebian:11 |
All versions, plus listed versions 2.3.1+dfsg-32.3.1+dfsg2-1 |
None known |
kerasPyPI |
before 3.15.0, plus listed versions 0.2.00.3.00.3.10.3.20.3.3 Show 45 moreShow less1.0.01.0.11.0.21.0.31.0.41.0.51.0.61.0.71.0.81.1.01.1.11.1.21.2.01.2.11.2.22.0.02.0.12.0.22.0.32.0.42.0.52.0.62.0.72.0.82.0.92.1.02.1.12.1.22.1.32.1.42.1.52.1.62.10.02.10.0rc02.10.0rc12.11.02.11.0rc02.11.0rc12.11.0rc22.11.0rc32.12.02.12.0rc02.12.0rc12.13.12.13.1rc0 |
3.15.0 |
python-KerasopenEuler:24.03-LTS-SP1 |
before 2.12.0-5.oe2403sp1 | 2.12.0-5.oe2403sp1 |
python-KerasopenEuler:24.03-LTS-SP3 |
before 2.12.0-5.oe2403sp3 | 2.12.0-5.oe2403sp3 |
python-KerasopenEuler:24.03-LTS-SP4 |
before 2.12.0-5.oe2403sp4 | 2.12.0-5.oe2403sp4 |
kerasUbuntu:18.04:LTS |
All versions, plus listed versions 1.0.7-22.1.1-1 |
None known |
kerasUbuntu:20.04:LTS |
All versions, plus listed versions 2.2.4-1 |
None known |
History
| Recorded | Changed | Sources |
|---|---|---|
| first indexed | CVE List, EUVD, GitHub Advisories, NVD, OSV |
Dates are when Every CVE recorded each change. Upstream dates may differ.
In short
CVE-2026-12570 is a medium vulnerability (CVSS 5.5) in Keras-team Keras-team/keras, published August 10, 2026. No source lists public exploit code. EPSS puts the chance of exploitation in the next 30 days at 0.19%, the 8th percentile. Fixed in keras 3.15.0.
References
huntr.com/bounties/a064f475-780a-409a-82f7-678512f27ad8
github.com/keras-team/keras/commit/4933ea4a5b3fcc24ceacdc276f5bb5dfbd06756c
nvd.nist.gov/vuln/detail/CVE-2026-12570
Show 9 moreShow less
security-tracker.debian.org/tracker/CVE-2026-12570
openeuler.org/zh/security/security-bulletins/detail?id=openEuler-SA-2026-3952
pypi.org/project/keras
ubuntu.com/security/CVE-2026-12570
cve.org/CVERecord?id=CVE-2026-12570
Left out here: 72 versions. The JSON record has them.