Impact
It is possible for a remote unauthenticated user to escape from the HQL execution context and perform a blind SQL injection to execute arbitrary SQL statements on the database backend, including when "Prevent unregistered users from viewing pages, regardless of the page rights" and "Prevent unregistered users from editing pages, regardless of the page rights" options are enabled.
Depending on the used database backend, the attacker may be able to not only obtain confidential information such as password hashes from the database, but also execute UPDATE/INSERT/DELETE queries.
The vulnerability may be tested in a default installation of XWIki Standard Flavor, including using the official Docker containers.
An example query, which leads to SQL injection with MySQL/MariaDB backend is shown below:
time curl "http://127.0.0.1:8080/rest/wikis/xwiki/query?q=where%20doc.name=length('a')org.apache.logging.log4j.util.Chars.SPACE%20or%201%3C%3E%271%5C%27%27%20union%20select%201,2,3,sleep(10)%20%23%27&type=hql&distinct=0"
When executed, the response from the server will come after a delay of 10 extra seconds, indicating successful execution of the injected SQL statement.
An example of a query for the PostgreSQL database backend is shown below:
curl "https://127.0.0.1:8080/rest/wikis/xwiki/query?q=where%20%24%24='%24%24=concat(%20chr(%2061%20),(chr(%2039%20))%20)%20;select%201%20--%20comment'&type=hql&distinct=0"
Both requests employ database backend dependent techniques of breaking out of HQL query context, described, for example, here: https://www.sonarsource.com/blog/exploiting-hibernate-injections.
Patches
This has been patched in 16.10.1, 16.4.6 and 15.10.16.
Workarounds
There is no known workaround, other than upgrading XWiki.
References
https://jira.xwiki.org/browse/XWIKI-22691
For more information
If you have any questions or comments about this advisory: Open an issue in Jira XWiki.org Email us at Security Mailing List
Attribution
Sergey Anufrienko from Kaspersky ICS-CERT vulnerability research team reported this vulnerability.
Impact
NOTE: The Realtime WYSIWYG Editor extension was experimental, and thus not recommended, in the versions affected by this vulnerability. It has become enabled by default, and thus recommended, starting with XWiki 16.9.0.
A user with only edit right can join a realtime editing session where others, that where already there or that may join later, have script or programming access rights. This user can then insert script rendering macros that are executed for those users in the realtime session that have script or programming rights. The inserted scripts can be used to gain more access rights.
Here's an example that works with XWiki 15.10.9+ and 16.2.0+:
the attacker starts editing a wiki page in realtime (for which they have edit right) another user, with script or programming access right joins the editing session (e.g. by clicking on a link / URL provided by the attacker) the attacker inserts a script rendering macro, say {{velocity}}I can run scripts{{/velocity}}, in the edited content, using the WYSIWYG editor UI the edited content is reloaded for both the attacker and the other user, in order to render the inserted macro the attacker gets a rendering error message the other user sees "I can run scripts"
The attacker can obviously use more advanced scripts to gain access rights.
Before XWiki 15.10.9 and 16.2.0 the edited content was not re-rendered for all the users in the editing sesesion, but only for the user that inserted the macro. This means that in order to reproduce the problem the other user had to insert or update a macro or save and view the content.
Patches
This vulnerability has been patched in XWiki 15.10.12, 16.4.1 and 16.6.0-rc-1.
Workarounds
To avoid this vulnerability you can:
either disable the realtime WYSIWYG editing by disabling the xwiki-realtime CKEditor plugin from the WYSIWYG editor administration section or uninstall the Realtime WYSIWYG Editor extension (org.xwiki.platform:xwiki-platform-realtime-wysiwyg-ui)
References
https://jira.xwiki.org/browse/XWIKI-21949
For more information
If you have any questions or comments about this advisory:
Open an issue in Jira XWiki.org Email us at Security Mailing List
Impact
It is possible for a user with SCRIPT right to escape from the HQL execution context and perform a blind SQL injection to execute arbitrary SQL statements on the database backend.
Depending on the used database backend, the attacker may be able to not only obtain confidential information such as password hashes from the database, but also execute UPDATE/INSERT/DELETE queries.
The vulnerability may be tested in a default installation of XWIki Standard Flavor, including using the official Docker containers.
For example, with a MySQL or MariaDB database, you can use the following script (which a user having SCRIPT right but not PROGRAMMING right) to get the content of the xwikistrings table (which contain all the short string fields stored in objects, including passwords):
{{velocity}} $services.query.hql("where 1<>'1\'' union select concat(XWSNAME, XWSVALUE) from xwikistrings #'").execute() {{/velocity}}
Patches
This has been patched in 16.10.1, 16.4.6 and 15.10.16.
Workarounds
There is no known workaround, other than upgrading XWiki.
The protection added to this REST API is the same as the one used to validate complete select queries, making it more consistent. However, while the script API always had this protection for complete queries, it's important to note that it's a very strict protection and some valid, but complex, queries might suddenly require the author to have programming right.
References
https://jira.xwiki.org/browse/XWIKI-22718
For more information
If you have any questions or comments about this advisory: Open an issue in Jira XWiki.org Email us at Security Mailing List