CVE-2021-29579: Heap buffer overflow in `MaxPoolGrad`

Impact The implementation of tf.rawops.MaxPoolGrad is vulnerable to a heap buffer overflow: python import tensorflow as tf

originput = tf.constant([0.0], shape=[1, 1, 1, 1], dtype=tf.float32) origoutput = tf.constant([0.0], shape=[1, 1, 1, 1], dtype=tf.float32) grad = tf.constant([], shape=[0, 0, 0, 0], dtype=tf.float32) ksize = [1, 1, 1, 1] strides = [1, 1, 1, 1] padding = "SAME"

tf.rawops.MaxPoolGrad( originput=originput, origoutput=origoutput, grad=grad, ksize=ksize, strides=strides, padding=padding, explicitpaddings=[])

The implementation fails to validate that indices used to access elements of input/output arrays are valid:

cc for (int index = outstart; index < outend; ++index) { int inputbackpropindex = outargmaxflat(index); FastBoundsCheck(inputbackpropindex - instart, inend - instart); inputbackpropflat(inputbackpropindex) += outbackpropflat(index); }

Whereas accesses to inputbackpropflat are guarded by FastBoundsCheck, the indexing in outbackpropflat can result in OOB access.

Patches We have patched the issue in GitHub commit a74768f8e4efbda4def9f16ee7e13cf3922ac5f7.

The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2, TensorFlow 2.3.3, TensorFlow 2.2.3 and TensorFlow 2.1.4, 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 by Ying Wang and Yakun Zhang of Baidu X-Team.