Skip to content

Latest commit

 

History

History
26 lines (18 loc) · 1.35 KB

tfsa-2021-196.md

File metadata and controls

26 lines (18 loc) · 1.35 KB
Raw

TFSA-2021-196: Overflow/crash in tf.tile when tiling tensor is large

CVE Number

CVE-2021-41198

Impact

If tf.tile is called with a large input argument then the TensorFlow process will crash due to a CHECK-failure caused by an overflow.

import tensorflow as tf
import numpy as np
tf.keras.backend.tile(x=np.ones((1,1,1)), n=[100000000,100000000, 100000000])

The number of elements in the output tensor is too much for the int64_t type and the overflow is detected via a CHECK statement. This aborts the process.

Patches

We have patched the issue in GitHub commit 9294094df6fea79271778eb7e7ae1bad8b5ef98f (merging #51138).

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.