CVE-2020-26266: Uninitialized memory access in Eigen types in TensorFlow

Severity: medium (CVSS 5.3 v3.1)
Exploited in the wild: not listed in CISA KEV
EPSS: 0.3% probability of exploitation in the next 30 days
Public exploit or detection template: yes
Published: 2020-12-10 · updated 2026-06-17
Weakness: CWE-908

Description

In affected versions of TensorFlow under certain cases a saved model can trigger use of uninitialized values during code execution. This is caused by having tensor buffers be filled with the default value of the type but forgetting to default initialize the quantized floating point types in Eigen. This is fixed in versions 1.15.5, 2.0.4, 2.1.3, 2.2.2, 2.3.2, and 2.4.0.

Affected products

tensorflow: before 1.15.5; from 2.0.0 before 2.0.4; from 2.1.0 before 2.1.3; from 2.2.0 before 2.2.2; from 2.3.0 before 2.3.2 — fixed in 1.15.5, 2.0.4, 2.1.3, 2.2.2, 2.3.2

How to fix

Upgrade to a fixed version: tensorflow 1.15.5, 2.0.4, 2.1.3, 2.2.2. Check the vendor advisory for mitigations.

References

Get alerted when CVE-2020-26266 or the next vulnerability affects your stack