Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Wed, 3 Jun 2015 20:04:25 +1200
From: Matthew Daley <mattd@...fuzz.com>
To: oss-security@...ts.openwall.com, cve-assign@...re.org
Subject: CVE requests / Advisory: Codestyling Localization (Wordpress plugin)
 - multiple RCE via CSRF, multiple XSS

Affected software: Codestyling Localization (Wordpress plugin)
Affected versions: ? -> 1.99.30
Website: http://www.code-styling.de/english/development/wordpress-plugin-codestyling-localization-en
Website: https://wordpress.org/plugins/codestyling-localization/ (now gone)
Reported by: Matthew Daley

I'd like to request CVE IDs for these issues. This is the first such
request; this message serves as an advisory as well.

I have received no response from the plugin author since 2015-04-03,
hence there are no vendor fixes for the below issues.

The plugin has since been removed from the Wordpress plugin directory,
so I don't have the ability to easily link to source code. FWIW, line
references in this advisory refer to what was the latest version of
the plugin, 1.99.30.


Description:

The plugin contains multiple AJAX actions that, while having the
necessary permission checks, do not have anti-CSRF protection (ie.
nonces). Hence, attackers can, with the usual victim interaction
that's required to trigger a CSRF attack, force logged-in victims with
the "manage_options" capability (ie. Super Admins or Administrators)
to perform a number of administrative actions.

Furthermore, some of these administrative actions use user-provided
input in an unsafe way and can be exploited in order to gain remote
code execution on the victim's Wordpress installation or to perform
reflected XSS attacks.


* Issue #1: RCE via CSRF in AJAX action "csp_po_scan_source_file":

This action is designed to add in translatable strings from one or
more PHP files into an existing Portable Object (PO) file (just as
`xgettext` would in 'join' mode).

The issue with this action is that the user-specified PO and PHP file
paths are not validated in any way. It is possible to abuse the action
by executing it with the following user-specified parameters:

- Set "pofile", the file to read and write as a PO-format file, to the
location of a writable, web-accessible PHP file on the system. For
example, "/var/www/wordpress/wp-includes/ID3/module.audio.dts.php".
(This file is unlikely to be executed often by WordPress in normal
functioning and so is a good choice for being overwritten entirely.)
- Set "php", the list of files to extract translatable strings from,
to include the location of a crafted file that contains a translatable
string with embedded PHP code in it. For example, the contents of this
file could be "<?php __('<?php echo `id` ?>') ?>", which would allow
the remote execution of the `id` command. Assuming that the
"allow_url_fopen" PHP configuration setting is enabled, this file can
be retrieved over a FTP connection to an attacker-controlled server,
which allows the file to contain whatever the attacker desires.
- Set "num" to 0 and "cnt" to 1 (indicating that 0 PHP files have been
processed so far and that there is a total of 1 to process)

Executing the action with these parameters will cause the file
specified in "pofile" (a writable, web-accessible PHP file) to be
overwritten with the PO-format output. One of the translatable strings
in this output will contain the PHP code, unescaped, from the file
specified in "php". Hence, the attacker then only needs to access the
overwritten "pofile" file remotely (by requesting it directly by URL)
for the attacker's code to then be executed by the server.

I have written a simple POC (not given here) to test this. It takes
the form of a Python script that hosts dummy FTP and HTTP servers that
serve the required payloads in order for this attack to function. The
attacker would then induce the victim, who is logged into the
Wordpress administration interface, to navigate to the script's HTTP
server. The payloads served by the script's HTTP server will cause the
victim to send a CSRF-forced request to the csp_po_scan_source_file
action with the necessary parameters (as described above). The
csp_po_scan_source_file action will connect back to the script's FTP
server in order to retrieve the PHP file to scan for translatable
strings, one of which contains the attacker-specified PHP code to be
remotely executed. Finally, the script watches the content of the
target file that is to be overwritten; when it sees that it has been,
it shows the result of the attacker-specified code's execution.

A sample run:

$ ./csrf-rce-csp_po_ajax_handle_scan_source_file.py
http://[redacted]/wordpress/ /var/www/wordpress id
[redacted] - - [03/Apr/2015 11:02:52] "GET / HTTP/1.1" 200 -
Pre-CSRF payload sent
[redacted] - - [03/Apr/2015 11:02:53] "GET /csrf HTTP/1.1" 200 -
CSRF payload sent
Input payload sent
uid=33(www-data) gid=33(www-data) groups=33(www-data)

The final line in the above output shows the script's result, namely
the output of running `id` on the victim's server.


* Issue #2: RCE via CSRF in AJAX action "csp_po_save_catalog_entry":

This action is similar in functioning to the previous action, except
that it takes its translated string inputs directly from
user-specified parameters instead of extracting them from a
user-specified file.

The action can therefore be exploited in a similar fashion, with one
caveat: the writable, web-accessible PHP file that is to be
overwritten with PO-format output ("path" and "file" POST parameters)
needs to appear to have at least one translatable string in it when it
is treated as a PO-format file. One of the ways that this can occur is
for a quotation-mark surrounded string literal to be alone on a single
line in the file (ignoring whitespace). Lines like this are parsed by
the PO-format parser as translatable strings. Hence, there will be at
least one translatable string extracted from the file that can then be
overwritten by the user-specified translatable string before it's
written out again to the writable, web-accessible PHP file.

I have written a simple POC (not given here) to test this. It takes
the form of an HTML file which sends the required POST parameters for
the vulnerability to be triggered. Once the victim has navigated to
the POC and the CSRF has been performed, the attacker can request the
target file remotely via URL and have their code executed, similarly
to the previous vulnerability.

(The benefit of this vulnerability is that the "allow_url_fopen" PHP
configuration setting does not need to be enabled in order for this
attack to function. However, the possible choices for the target file
are reduced; the file needs to appear to have a translatable string
when treated as a PO-format file, which the previous vulnerability
does not require.)


* Other issues: Multiple XSS in various AJAX actions

These XSS are all due to reflected unescaped POST parameters in
certain AJAX actions' responses. The actions and parameters are:

* Action "csp_self_protection_result": parameter "data[php][]"
* Action "csp_po_dlg_delete": parameters "language", "name",
"numlangs", "path", "row", "subpath"
* Action "csp_po_dlg_rescan": parameter "language"
* Action "csp_po_generate_mo_file": parameter "pofile"
* Action "csp_po_save_catalog_entry": parameter "file"
* Action "csp_po_scan_source_file": parameter "pofile"


- Matthew Daley

Powered by blists - more mailing lists

Your e-mail address:

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Powered by Openwall GNU/*/Linux - Powered by OpenVZ