CriticalVulnerability

CVE-2021-37678: TensorFlow arbitrary code execution when deserializing Keras model from YAML

Source: NVD(opens in a new tab)
Identifier: CVE-2021-37678
Published: August 12, 2021
Record updated: October 9, 2026

Affected: tensorflow < 2.3.4, fixed in 2.3.4, released August 10, 2021, 3 days before disclosure
tensorflow >= 2.4.0, < 2.4.3, fixed in 2.4.3, released August 10, 2021, 3 days before disclosure
tensorflow = 2.5.0, fixed in 2.5.1, released August 10, 2021, 3 days before disclosure
and 6 more
Known exploitation: Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
EPSS: 0.4%

Summary

TensorFlow and Keras can be tricked into arbitrary code execution when deserializing a Keras model from YAML format. The implementation uses `yaml.unsafe_load`, which can execute arbitrary code on the input. The YAML format support has been removed for now.

Mitigation

Patched in GitHub commit 23d6383eb6c14084a8fc3bdf164043b974818012. The fix will be included in TensorFlow 2.6.0, and the commit will also be cherry-picked to TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4.

Details

Scores and exploitation

CVSS score: 9.3(critical)
Patch available: Yes
EPSS: 0.4%Estimated probability that this CVE is exploited in the next 30 days, from the FIRST Exploit Prediction Scoring System. Read on November 4, 2026.

Weaknesses and techniques

Weakness (CWE): CWE-502
Attack pattern (CAPEC): CAPEC-586

Affected

Packages: tensorflow@< 2.3.4 (fixed: 2.3.4)
tensorflow@>= 2.4.0, < 2.4.3 (fixed: 2.4.3)
tensorflow@= 2.5.0 (fixed: 2.5.1)
tensorflow-cpu@< 2.3.4 (fixed: 2.3.4)
tensorflow-cpu@>= 2.4.0, < 2.4.3 (fixed: 2.4.3)
tensorflow-cpu@= 2.5.0 (fixed: 2.5.1)
tensorflow-gpu@< 2.3.4 (fixed: 2.3.4)
tensorflow-gpu@>= 2.4.0, < 2.4.3 (fixed: 2.4.3)
tensorflow-gpu@= 2.5.0 (fixed: 2.5.1)
Vendors: Google

Classification

Attack type: Other
Attack sophistication: Moderate
AI component targeted: Framework
Impact: Confidentiality,
Integrity,
Availability
Labels: security

How this record was produced

Summaries, mitigations and labels on this page are generated automatically from the source. The linked source is authoritative. The source gives this advisory a description and no title, so the title above is a headline generated from the source text.

Original source: https://nvd.nist.gov/vuln/detail/CVE-2021-37678

First tracked: February 15, 2026 at 08:39 PM ET, from NVD/CVE Database

Classifier: claude-haiku-5-5, prompt v4

Classifier confidence: 95% (self-assessed)

Affected packages in the Exposure Registry

Matched by package name and ecosystem. Each entry shows whether the package delegates to a language model and how many tracked packages depend on it.

tensorflowPyPINo LLM dependency detected · 26 tracked dependents
tensorflow-gpuPyPINo LLM dependency detected · 2 tracked dependents

Summary

Mitigation

Affected packages in the Exposure Registry

Related items