Deserialization in Google Keras
CVE-2025-8747
A safe mode bypass vulnerability in the `Model.load_model` method in Keras versions 3.0.0 through 3.10.0 allows an attacker to achieve arbitrary code execution by convincing a user to load a specially crafted `.keras` model archive.
Vulnerability class: Insecure Deserialization
Published · last modified .
CVSS v3 metric
CVSS v3 base score 7.8 (High). Vector: CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H.
EPSS exploit prediction
EPSS: 0.001 (1.5th percentile), scored .
Very low probability of exploitation in the next 30 days; routine patching cadence is appropriate. 2nd percentile — 1.5% of CVEs in the catalogue have a lower EPSS than this one. How to read EPSS.
EPSS over last 30 days · oldest: 0.001 · newest: 0.001 · change: -0.000
Affected products
- Google Keras — versions 3.0.0
- Keras
Weakness classification (CWE)
Public proof-of-concept exploits
References
- github.com/keras-team/keras/pull/21429 (patch, Issue Tracking)
- jfrog.com/blog/keras-safe_mode-bypass-vulnerability/ (Third Party Advisory, third-party-advisory)
Frequently asked questions
- What is CVE-2025-8747?
- CVE-2025-8747 is a high-severity vulnerability in Google Keras, classified under Deserialization of Untrusted Data. CVSS score: 7.8/10. Published 2025-08-11.
- How severe is CVE-2025-8747?
- High severity. CVSS v3 base score is 7.8 out of 10.
- Is CVE-2025-8747 known to be exploited?
- 1 public proof-of-concept repositories are indexed. Not currently listed in the CISA KEV catalog.