PT-2026-59857 · Pypi · Tensorflow

Published

2026-07-09

·

Updated

2026-07-09

CVSS v3.1

6.5

Medium

VectorAV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H

Impact

The [estimator for the cost of some convolution operations](https://github.com/tensorflow/tensorflow/blob/ffa202a17ab7a4a10182b746d230ea66f021fe16/tensorflow/core/grappler/costs/op level cost estimator.cc#L189-L198) can be made to execute a division by 0:
python
import tensorflow as tf

@tf.function
def test():
 y=tf.raw ops.AvgPoolGrad(
  orig input shape=[1,1,1,1],
  grad=[[[[1.0],[1.0],[1.0]]],[[[2.0],[2.0],[2.0]]],[[[3.0],[3.0],[3.0]]]],
  ksize=[1,1,1,1],
  strides=[1,1,1,0],
  padding='VALID',
  data format='NCHW')
 return y

test()
The function fails to check that the stride argument is stricly positive:
cc
int64 t GetOutputSize(const int64 t input, const int64 t filter,
           const int64 t stride, const Padding& padding) {
 // Logic for calculating output shape is from GetWindowedOutputSizeVerbose() 
 // function in third party/tensorflow/core/framework/common shape fns.cc.
 if (padding == Padding::VALID) {
  return (input - filter + stride) / stride;
 } else { // SAME.
  return (input + stride - 1) / stride;
 }
} 
Hence, the fix is to add a check for the stride argument to ensure it is valid.

Patches

We have patched the issue in GitHub commit 3218043d6d3a019756607643cf65574fbfef5d7a.
The 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.

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 by Yu Tian of Qihoo 360 AIVul Team.

Fix

Found an issue in the description? Have something to add? Feel free to write us 👾

Related Identifiers

PYSEC-2026-3242

Affected Products

Tensorflow