CVE-2021-37679: TensorFlow - Uninitialized Memory Leak in tf.map_fn
Summary
CVE-2021-37679: TensorFlow is an end-to-end open source platform for machine learning. Affects tensorflow across 3 version ranges.
>= 2.3.0, < 2.3.4
>= 2.4.0, < 2.4.3
< 4e2565483d0ffcadc719bd44893fb7f609bb5f12
Fixed in 2.3.4 or 2.4.3, depending on the release line you run. How to fix ↓
Also tracked for
- tensorflow · Bitnami
- tensorflow-cpu · PyPI
- tensorflow-gpu · PyPI
The same CVE-2021-37679 is tracked separately for each ecosystem and vendor. Its fix information is the same; the package and affected versions differ.
Check if this vulnerability affects your application
Upload a manifest or SBOM and we'll show whether you're affected, plus every other risky package in the same scan.
No signup required · Results in about 60 seconds
Background
TensorFlow is an end-to-end open source platform for machine learning. It provides a comprehensive ecosystem of tools, libraries, and resources for building and deploying machine learning models. TensorFlow is widely used in various domains, including research, industry, and academia. This vulnerability affects versions of TensorFlow that support the tf.map_fn function when nested within another tf.map_fn call.
Vulnerability Detail
The vulnerability arises when a RaggedTensor is used as the input tensor for nested tf.map_fn calls without providing a function signature. In this scenario, the code assumes that the output is a fully specified tensor and fills the output buffer with uninitialized contents from the heap. As a result, the t and z outputs are not identical, and the last row of t contains data from the heap. An attacker can exploit this to leak other memory information. The bug is caused by the conversion from a Variant tensor to a RaggedTensor, where the implementation fails to check that all inner shapes match, leading to additional dimensions. This vulnerability can also result in data loss if the input tensor is manipulated.
How To Fix
Latest published version: 2.21.0 as of 22 Oct 2026
pip install --upgrade "tensorflow==2.21.0"Upgrading to it may include breaking changes; the versions below are the minimum fixed release for this vulnerability.
If you can't move to the latest
The minimum fixed version depends on the release line you run.
-
>= 2.3.0, < 2.3.42.3.4pip install --upgrade "tensorflow==2.3.4" -
>= 2.4.0, < 2.4.32.4.3pip install --upgrade "tensorflow==2.4.3"
Tip — Not sure which release line you're on? Vulert can find the fix for the version you are using. Scan free →
Workaround
As a workaround, users can avoid nesting tf.map_fn calls with RaggedTensor inputs or provide a function signature when using RaggedTensor as the input tensor. This can help prevent the uninitialized memory leak and potential data loss.
Conclusion
The vulnerability in TensorFlow that allows uninitialized memory leaks in nested tf.map_fn calls with RaggedTensor inputs has been addressed by the TensorFlow team. Users are advised to upgrade to the patched versions to mitigate the risk. For further support on vulnerability remediation, please contact DevNack.com.
See every vulnerability recorded for tensorflow →
FAQ
TensorFlow is an open source platform for machine learning that provides a comprehensive set of tools and resources for building and deploying machine learning models.
This vulnerability allows an attacker to leak memory information and potentially manipulate data in TensorFlow models.
The vulnerability affects TensorFlow versions 2.3.0 and above, including TensorFlow 2.6.0.
To fix the vulnerability, users should upgrade to one of the patched versions: TensorFlow 2.6.0, TensorFlow 2.5.1, TensorFlow 2.4.3, or TensorFlow 2.3.4.
Yes, users can avoid nesting tf.map_fn calls with RaggedTensor inputs or provide a function signature when using RaggedTensor as the input tensor.
References
Related vulnerabilities
Recent vulnerabilities recorded for tensorflow.
