CVE-2021-41197: Crashes due to overflow and `CHECK`-fail in ops with large tensor shapes
Impact 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 int64t. 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).
Patches 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.
For more information Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
Attribution This vulnerability has been reported externally via GitHub issue, GitHub issue and GitHub issue.
Other sources
TensorFlow is an open source platform for machine learning. In affected versions 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 int64t. 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. This is similar to CVE-2021-29584. 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.
Affected Software
Remediation
Event History
Frequently Asked Questions
What is the severity of CVE-2021-41197?
CVE-2021-41197 has a moderate severity level due to potential integer overflow issues.
How do I fix CVE-2021-41197?
To fix CVE-2021-41197, upgrade TensorFlow to version 2.4.4 or later.
Which TensorFlow versions are affected by CVE-2021-41197?
Versions prior to TensorFlow 2.4.4, including older 2.5.x and 2.6.x versions, are affected by CVE-2021-41197.
What vulnerability does CVE-2021-41197 exploit?
CVE-2021-41197 exploits integer overflow errors in tensor multiplication without overflow checks.
Are both TensorFlow CPU and GPU versions affected by CVE-2021-41197?
Yes, both TensorFlow CPU and GPU versions are affected by CVE-2021-41197.