CVE-2020-26270: CHECK-fail in LSTM with zero-length input in TensorFlow

Published Dec 10, 2020
·
Updated

Impact Running an LSTM/GRU model where the LSTM/GRU layer receives an input with zero-length results in a CHECK failure when using the CUDA backend.

This can result in a query-of-death vulnerability, via denial of service, if users can control the input to the layer.

Patches We have patched the issue in GitHub commit 14755416e364f17fb1870882fa778c7fec7f16e3 and will release TensorFlow 2.4.0 containing the patch. TensorFlow nightly packages after this commit will also have the issue resolved.

Since this issue also impacts TF versions before 2.4, we will patch all releases between 1.15 and 2.3 inclusive.

For more information Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.

Other sources

In affected versions of TensorFlow running an LSTM/GRU model where the LSTM/GRU layer receives an input with zero-length results in a CHECK failure when using the CUDA backend. This can result in a query-of-death vulnerability, via denial of service, if users can control the input to the layer. 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 Software

20 affected componentsFixes available
pip/tensorflow-gpu>=2.3.0<2.3.2
2.3.2
pip/tensorflow-gpu>=2.2.0<2.2.2
2.2.2
pip/tensorflow-gpu>=2.1.0<2.1.3
2.1.3
pip/tensorflow-gpu>=2.0.0<2.0.4
2.0.4
pip/tensorflow-gpu<1.15.5
1.15.5
pip/tensorflow-cpu>=2.3.0<2.3.2
2.3.2
pip/tensorflow-cpu>=2.2.0<2.2.2
2.2.2
pip/tensorflow-cpu>=2.1.0<2.1.3
2.1.3
pip/tensorflow-cpu>=2.0.0<2.0.4
2.0.4
pip/tensorflow-cpu<1.15.5
1.15.5
pip/tensorflow>=2.3.0<2.3.2
2.3.2
pip/tensorflow>=2.2.0<2.2.2
2.2.2
pip/tensorflow>=2.1.0<2.1.3
2.1.3
pip/tensorflow>=2.0.0<2.0.4
2.0.4
pip/tensorflow<1.15.5
1.15.5
Google TensorFlow<1.15.5
Google TensorFlow>=2.0.0<2.0.4
Google TensorFlow>=2.1.0<2.1.3
Google TensorFlow>=2.2.0<2.2.2
Google TensorFlow>=2.3.0<2.3.2

Remediation

Patch Available

https://github.com/tensorflow/tensorflow/commit/14755416e364f17fb1870882fa778c7fec7f16e3

Patch Available

https://github.com/tensorflow/tensorflow/security/advisories/GHSA-m648-33qf-v3gp

Event History

Dec 10, 2020
Advisory Published
via GitHub·07:07 PM
CVE Published
via MITRE·10:10 PM
Data Sourced
via MITRE·10:10 PM
DescriptionSeverityWeakness
Data Sourced
via NVD·11:15 PM
RemedyDescriptionSeverityWeaknessAffected Software

Frequently Asked Questions

1

What is the impact of CVE-2020-26270?

CVE-2020-26270 can lead to a denial of service via a query-of-death vulnerability when an LSTM/GRU model receives zero-length input.

2

How do I fix CVE-2020-26270?

To fix CVE-2020-26270, upgrade TensorFlow to versions 1.15.5, 2.0.4, 2.1.3, 2.2.2, or 2.3.2.

3

Which versions of TensorFlow are affected by CVE-2020-26270?

CVE-2020-26270 affects TensorFlow versions prior to 1.15.5 and between 2.0.0 to 2.3.0.

4

Is there a patch available for CVE-2020-26270?

Yes, the patch for CVE-2020-26270 is included in TensorFlow versions 1.15.5 and 2.0.4 through 2.3.2.

5

How can CVE-2020-26270 lead to a denial of service?

CVE-2020-26270 can lead to a denial of service by causing a CHECK failure when the LSTM/GRU layer processes zero-length input.

Company

Resources

Contact

SecAlerts Pty Ltd.
132 Wickham Terrace
Fortitude Valley,
QLD 4006, Australia
info@secalerts.co
By using SecAlerts services, you agree to our services end-user license agreement. This website is safeguarded by reCAPTCHA and governed by the Google Privacy Policy and Terms of Service. All names, logos, and brands of products are owned by their respective owners, and any usage of these names, logos, and brands for identification purposes only does not imply endorsement. If you possess any content that requires removal, please get in touch with us.
© 2026 SecAlerts Pty Ltd.
ABN: 70 645 966 203, ACN: 645 966 203