First published: Fri Sep 16 2022(Updated: )
TensorFlow is an open source platform for machine learning. When `Conv2DBackpropInput` receives empty `out_backprop` inputs (e.g. `[3, 1, 0, 1]`), the current CPU/GPU kernels `CHECK` fail (one with dnnl, the other with cudnn). This can be used to trigger a denial of service attack. We have patched the issue in GitHub commit 27a65a43cf763897fecfa5cdb5cc653fc5dd0346. 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. There are no known workarounds for this issue.
Credit: security-advisories@github.com
Affected Software | Affected Version | How to fix |
---|---|---|
TensorFlow Keras | <2.7.2 | |
TensorFlow Keras | >=2.8.0<2.8.1 | |
TensorFlow Keras | >=2.9.0<2.9.1 | |
TensorFlow Keras | =2.10-rc0 | |
TensorFlow Keras | =2.10-rc1 | |
TensorFlow Keras | =2.10-rc2 | |
TensorFlow Keras | =2.10-rc3 |
Sign up to SecAlerts for real-time vulnerability data matched to your software, aggregated from hundreds of sources.
CVE-2022-35999 is classified as a denial of service vulnerability that can lead to application crashes.
To remediate CVE-2022-35999, upgrade TensorFlow to a version that is not vulnerable, specifically versions later than 2.10-rc3.
CVE-2022-35999 affects TensorFlow versions up to 2.10-rc3, including versions 2.7.2, 2.8.0, and 2.9.0 through 2.9.1.
CVE-2022-35999 affects the Conv2DBackpropInput operation within TensorFlow when it receives empty out_backprop inputs.
Yes, CVE-2022-35999 can be exploited remotely to trigger a denial of service condition.