CVE-2021-36749: Apache Druid: The HTTP inputSource allows authenticated users to read data from other sources than intended (incomplete fix of CVE-2021-26920)

8 views
Skip to first unread message

Clint Wylie

unread,
Sep 23, 2021, 7:02:58 PMSep 23
to anno...@apache.org, d...@druid.apache.org, druid...@googlegroups.com
Severity: low

Description:

In the Druid ingestion system, the InputSource is used for reading
data from a certain data source. However, the HTTP InputSource allows
authenticated users to read data from other sources than intended,
such as the local file system, with the privileges of the Druid server
process. This is not an elevation of privilege when users access Druid
directly, since Druid also provides the Local InputSource, which
allows the same level of access. But it is problematic when users
interact with Druid indirectly through an application that allows
users to specify the HTTP InputSource, but not the Local InputSource.
In this case, users could bypass the application-level restriction by
passing a file URL to the HTTP InputSource.

This issue was previously mentioned as being fixed in 0.21.0 as per
CVE-2021-26920 but was not fixed in 0.21.0 or 0.21.1.

Mitigation:

Users can avoid the issue by upgrading to 0.22.0 or a higher version.

In an earlier version than 0.22.0, when the user application wants to
restrict the access to the local file system, it should disallow all
InputSources that can read local files, that is the Local, HTTP, and
HDFS InputSources.

Credit:

This issue was originally discovered by chybeta from the Security Team
of Alibaba Cloud.
ABKing and g0udan from the Security Team of Xiaomi discovered that it
was still an issue after CVE-2021-26920.
Reply all
Reply to author
Forward
0 new messages