tfsa-2021-199.md

TFSA-2021-199: Crash in max_pool3d when size argument is 0 or negative

CVE Number

CVE-2021-41196

Impact

The Keras pooling layers can trigger a segfault if the size of the pool is 0 or if a dimension is negative:

import tensorflow as tf

pool_size = [2, 2, 0]
layer = tf.keras.layers.MaxPooling3D(strides=1, pool_size=pool_size)
input_tensor = tf.random.uniform([3, 4, 10, 11, 12], dtype=tf.float32)
res = layer(input_tensor)

This is due to the TensorFlow's implementation of pooling operations where the values in the sliding window are not checked to be strictly positive.

Patches

We have patched the issue in GitHub commit 12b1ff82b3f26ff8de17e58703231d5a02ef1b8b (merging #51975).

The fix will be included in TensorFlow 2.7.0. We will also cherrypick this commit on TensorFlow 2.6.1, TensorFlow 2.5.2, and TensorFlow 2.4.4, 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 externally via a GitHub issue.