Skip to content

Seg fault in `ndarray_tensor_bridge` due to zero and large inputs

Low
pak-laura published GHSA-jq6x-99hj-q636 Nov 18, 2022

Package

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

Affected versions

< 2.11.0

Patched versions

2.8.4, 2.9.3, 2.10.1, 2.11.0

Description

Impact

If a numpy array is created with a shape such that one element is zero and the others sum to a large number, an error will be raised. E.g. the following raises an error:

np.ones((0, 2**31, 2**31))

An example of a proof of concept:

import numpy as np
import tensorflow as tf

input_val = tf.constant([1])
shape_val = np.array([i for i in range(21)])

tf.broadcast_to(input=input_val,shape=shape_val)

The return value of PyArray_SimpleNewFromData, which returns null on such shapes, is not checked.

Patches

We have patched the issue in GitHub commit 2b56169c16e375c521a3bc8ea658811cc0793784.

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 Pattarakrit Rattanukul.

Severity

Low

CVE ID

CVE-2022-41884

Weaknesses

No CWEs