CVE-2021-41197
TensorFlow allows tensor to have a large number of dimensions and each dimension can be as large as desired. However, the total number of elements in a tensor must fit within an int64_t. If an overflow occurs, MultiplyWithoutOverflow would return a negative result. In the majority of TensorFlow codebase this then results in a CHECK-failure. Newer constructs exist which return a Status instead of crashing the binary.
For example AddDim calls should be replaced by AddDimWithStatus.
This is similar to CVE-2021-29584 (and similar other reported vulnerabilities in TensorFlow, localized to specific APIs).
We have patched the issue in GitHub commits 7c1692bd417eb4f9b33ead749a41166d6080af85 (merging #51732), d81b1351da3e8c884ff836b64458d94e4a157c15 (merging #51717), a871989d7b6c18cdebf2fb4f0e5c5b62fbc19edf (merging #51658), and d81b1351da3e8c884ff836b64458d94e4a157c15 (merging #51973). It is possible that other similar instances exist in TensorFlow, we will issue fixes as these are discovered.
The fix will be included in TensorFlow 2.7.0. We will also cherrypick this commit on TensorFlow 2.6.1, TensorFlow 2.5.2, and TensorFlow 2.4.4, as these are also affected and still in supported range.
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
This vulnerability has been reported externally via GitHub issue, GitHub issue and GitHub issue.