GHSA-x563-6hqv-26mr: Ibis PyArrow dependency allows arbitrary code execution when loading a malicious data file

Impact

Deserialization of untrusted data in IPC and Parquet readers in PyArrow versions 0.14.0 to 14.0.0 allows arbitrary code execution. An application is vulnerable if it reads Arrow IPC, Feather or Parquet data from untrusted sources (for example user-supplied input files). This vulnerability only affects PyArrow, not other Apache Arrow implementations or bindings.

Note that Ibis itself makes extremely limited use of pyarrow.parquet.read_table:

1. read_table is used in tests, where the input file is entirely controlled by the Ibis developers
2. read_table is used in the ibis/examples/init.py as a fallback for backends that don’t support reading Parquet directly. Parquet data used in ibis.examples are also managed by the Ibis developers. This Parquet data is generated from CSV files and SQLite databases.
3. The Pandas and Dask backends both use PyArrow to read Parquet files and are therefore affected.

Ibis does not make use of APIs that directly read from either Arrow IPC files or Feather files.

Patches

Ibis imports the pyarrow_hotfix package wherever pyarrow is used, as of version 7.1.0.

Upgrading to Arrow 14.0.1 is also a possible solution, starting in Ibis 7.1.0.

Workarounds

Install pyarrow_hotfix and run import pyarrow_hotfix ahead of any and all import ibis statements.

For example:

becomes

import pyarrow_hotfix import ibis

References

https://www.cve.org/CVERecord?id=CVE-2023-47248
https://nvd.nist.gov/vuln/detail/CVE-2023-47248

References

• GHSA-x563-6hqv-26mr
• ibis-project/ibis@0fa1e5d
• https://github.com/ibis-project/ibis/releases/tag/7.1.0