Missing validation causes denial of service via `StagePeek`

Overview

Impact

The implementation of tf.raw_ops.StagePeek does not fully validate the input arguments. This results in a CHECK-failure which can be used to trigger a denial of service attack:

import tensorflow as tf

index = tf.constant([], shape=[0], dtype=tf.int32)
tf.raw_ops.StagePeek(index=index, dtypes=[tf.int32])

The code assumes index is a scalar but there is no validation for this before accessing its value:

std::size_t index = ctx->input(0).scalar<int>()();

Patches

We have patched the issue in GitHub commit cebe3c45d76357d201c65bdbbf0dbe6e8a63bbdb.

The fix will be included in TensorFlow 2.9.0. We will also cherrypick this commit on TensorFlow 2.8.1, TensorFlow 2.7.2, and TensorFlow 2.6.4, 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 Neophytos Christou from Secure Systems Lab at Brown University.

Affected packages

• tensorflow < 2.6.4
• tensorflow >= 2.7.0, < 2.7.2
• tensorflow >= 2.8.0, < 2.8.1
• tensorflow-cpu < 2.6.4
• tensorflow-cpu >= 2.7.0, < 2.7.2
• tensorflow-cpu >= 2.8.0, < 2.8.1
• tensorflow-gpu < 2.6.4
• tensorflow-gpu >= 2.7.0, < 2.7.2
• tensorflow-gpu >= 2.8.0, < 2.8.1

Remediation

Upgrade to a patched release:

• tensorflow 2.6.4
• tensorflow 2.7.2
• tensorflow 2.8.1
• tensorflow-cpu 2.6.4
• tensorflow-cpu 2.7.2
• tensorflow-cpu 2.8.1
• tensorflow-gpu 2.6.4
• tensorflow-gpu 2.7.2
• tensorflow-gpu 2.8.1