CVE-2025-49655GHSA-cvhh-q5g5-qprpPYSEC-2026-368
Keras framework vulnerable to deserialization of untrusted data
Keras keraspipLibraries & packages
CVSS
9.8CriticalExploitationPublic exploit codeCISA's assessment says proof of concept code is public, assessed 17 Oct 2025. Not in the KEV catalogCISA via CVE List
Exploit odds (EPSS)0.75%Chance of exploitation activity in the next 30 days, higher than 54% of CVEs. Scored 27 Oct 2026FIRST EPSS
What fixes it
A fixed version on one release line is not the upgrade target for every line. Check your exact version and any vendor backports.
Description
Deserialization of untrusted data can occur in versions of the Keras framework running versions 3.11.0 up to but not including 3.11.3, enabling a maliciously uploaded Keras file containing a TorchModuleWrapper class to run arbitrary code on an end user’s system when loaded despite safe mode being enabled. The vulnerability can be triggered through both local and remote files.
Text from CVE List
Affected versions 6
- keras, pip>= 3.11.0, < 3.11.3GitHub
- Kerasaffected: 3.11.0 before 3.11.3CVE List
- keras, PyPIfrom 3.11.0, fixed in 3.11.3OSV
- keras, PyPI3.11.0OSV
- keras, PyPI3.11.1OSV
- keras, PyPI3.11.2OSV
References 7
- hiddenlayer.com/sai_security_advisor/2025-10-kerasAdvisoryvia GitHub
- github.com/pypa/advisory-database/tree/main/vulns/keras/PYSEC-2026-368.yamlAdvisoryvia GitHub
- github.com/advisories/GHSA-cvhh-q5g5-qprpAdvisoryvia GitHub
- hiddenlayer.com/sai_security_advisor/2025-10-keras/Advisoryvia CVE List
- nvd.nist.gov/vuln/detail/CVE-2025-49655Referencevia GitHub
- github.com/keras-team/keras/pull/21575Referencevia CVE List
- pypi.org/project/kerasReferencevia GitHub