Skip to content
TensorFlowGHSA-g468-qj8g-vcjc

TensorFlow vulnerable to `CHECK`-fail in `tensorflow::full_type::SubstituteFromAttrs`

LowCVE-2022-36016 · Published Sep 16, 2022 · updated Jul 13, 2026

### Impact When [`tensorflow::full_type::SubstituteFromAttrs`](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/ops/math_ops.cc) receives a `FullTypeDef& t` that is not exactly three args, it triggers a `CHECK`-fail instead of returning a status. ```cpp Status SubstituteForEach(AttrMap& attrs, FullTypeDef& t) { DCHECK_EQ(t.args_size(), 3); const auto& cont = t.args(0); const auto& tmpl = t.args(1); const auto& t_var = t.args(2); ``` ### Patches We have patched the issue in GitHub commit [6104f0d4091c260ce9352f9155f7e9b725eab012](https://github.com/tensorflow/tensorflow/commit/6104f0d4091c260ce9352f9155f7e9b725eab012). The fix will be included in TensorFlow 2.10.0. We will also cherrypick this commit on TensorFlow 2.9.1, TensorFlow 2.8.1, and TensorFlow 2.7.2, as these are also affected and still in supported range. ### For more information Please consult [our security guide](https://github.com/tensorflow/tensorflow/blob/master/SECURITY.md) for more information regarding the security model and how to contact us with issues and questions.

GitHub advisory

Affected versions

PackageAffectedFixed in
tensorflow
PyPI
< 2.7.22.7.2
>= 2.8.0, < 2.8.12.8.1
>= 2.9.0, < 2.9.12.9.1
Details and references

More TensorFlow advisories

All TensorFlow
Advisory
TensorFlow vulnerable to `CHECK` fail in `Save` and `SaveSlices`
Medium5.9Sep 16, 2022
TensorFlow vulnerable to `CHECK` fail in `ParameterizedTruncatedNormal`
Medium5.9Sep 16, 2022
TensorFlow vulnerable to `CHECK` fail in `LRNGrad`
Medium5.9Sep 16, 2022
TensorFlow vulnerable to segfault in `RaggedBincount`
Medium5.9Sep 16, 2022
TensorFlow vulnerable to `CHECK` fail in `tf.linalg.matrix_rank`
Medium5.9Sep 16, 2022
TensorFlow vulnerable to `CHECK` fail in `MaxPool`
Medium5.9Sep 16, 2022

Critical advisories by email

Wednesdays: the week’s critical and high advisories in the AI and data stack, with the fixed versions. Only in weeks that have some.

Double opt-in. Unsubscribe any time.