Revision as of 21:00, 29 January 2007

[Up]
OWASP Testing Guide v2 Table of Contents

Brief Summary

Web server fingerprinting is a critical task for the penetration tester. Knowing the version and type of a running web server allows testers to determine known vulnerabilities and the appropriate exploits to use during testing.

Description of the Issue

There are several different vendors and versions of web servers on the market today. Knowing the type of web server that you are testing significantly helps in the testing process, and will also change the course of the test. This information can be derived by sending the web server specific commands and analyzing the output, as each version of web server software may respond differently to these commands. By knowing how each type of web server responds to specific commands and keeping this information in a web server fingerprint database, a penetration tester can send these commands to the web server, analyze the respsonse, and compare it to the database of known signatures. Please note that it usually takes several different commands to accurately identify the web server, as different versions may react similarly to the same command. Rarely, however, do different versions react the same to all HTTP commands. So, by sending several different commands, you increase the accuracy of your guess.

Black Box testing and example

The simplest and most basic form of identifying a web server is to look at the Server field in the HTTP response header. For our experiments we use netcat. Consider the following HTTP Request-Response:

$ nc 202.41.76.251 80
HEAD / HTTP/1.0

HTTP/1.1 200 OK
Date: Mon, 16 Jun 2003 02:53:29 GMT
Server: Apache/1.3.3 (Unix)  (Red Hat/Linux)
Last-Modified: Wed, 07 Oct 1998 11:18:14 GMT
ETag: "1813-49b-361b4df6"
Accept-Ranges: bytes
Content-Length: 1179
Connection: close
Content-Type: text/html

$

from the Server field we understand that the server is Apache, version 1.3.3, running on Linux operating system. Three examples of the HTTP response headers are shown below:

From an Apache 1.3.23 server:

HTTP/1.1 200 OK 
Date: Sun, 15 Jun 2003 17:10: 49 GMT 
Server: Apache/1.3.23 
Last-Modified: Thu, 27 Feb 2003 03:48: 19 GMT 
ETag: 32417-c4-3e5d8a83 
Accept-Ranges: bytes 
Content-Length: 196 
Connection: close 
Content-Type: text/HTML 

From a Microsoft IIS 5.0 server:

HTTP/1.1 200 OK 
Server: Microsoft-IIS/5.0 
Expires: Yours, 17 Jun 2003 01:41: 33 GMT 
Date: Mon, 16 Jun 2003 01:41: 33 GMT 
Content-Type: text/HTML 
Accept-Ranges: bytes 
Last-Modified: Wed, 28 May 2003 15:32: 21 GMT 
ETag: b0aac0542e25c31: 89d 
Content-Length: 7369 

From a Netscape Enterprise 4.1 server:

HTTP/1.1 200 OK 
Server: Netscape-Enterprise/4.1 
Date: Mon, 16 Jun 2003 06:19: 04 GMT 
Content-type: text/HTML 
Last-modified: Wed, 31 Jul 2002 15:37: 56 GMT 
Content-length: 57 
Accept-ranges: bytes 
Connection: close 

From a SunONE 6.1 server:

HTTP/1.1 200 OK
Server: Sun-ONE-Web-Server/6.1
Date: Tue, 16 Jan 2007 14:53:45 GMT
Content-length: 1186
Content-type: text/html
Date: Tue, 16 Jan 2007 14:50:31 GMT
Last-Modified: Wed, 10 Jan 2007 09:58:26 GMT
Accept-Ranges: bytes
Connection: close

However, this testing methodology is not so good. There are several techniques that allow a web site to obfuscate or to modify the server banner string. For example we could obtain the following answer:

403 HTTP/1.1 
Forbidden Date: Mon, 16 Jun 2003 02:41: 27 GMT 
Server: Unknown-Webserver/1.0 
Connection: close 
Content-Type: text/HTML; 
charset=iso-8859-1 

In this case the server field of that response is obfuscated: we cannot know what type of web server is running.

Protocol behaviour

Refined techniques of testing take into consideration various characteristics of the several web servers available on the market. We will list some methodologies that allow us to deduce the type of web server in use.

HTTP header field ordering

The first method consists of observing the ordering of the several headers in the response. Every web server has an inner ordering of the header. We consider the following answers as an example:

Response from Apache 1.3.23

$ nc apache.example.com 80 
HEAD / HTTP/1.0 

HTTP/1.1 200 OK 
Date: Sun, 15 Jun 2003 17:10: 49 GMT 
Server: Apache/1.3.23 
Last-Modified: Thu, 27 Feb 2003 03:48: 19 GMT 
ETag: 32417-c4-3e5d8a83 
Accept-Ranges: bytes 
Content-Length: 196 
Connection: close 
Content-Type: text/HTML 

Response from IIS 5.0

$ nc iis.example.com 80 
HEAD / HTTP/1.0 

HTTP/1.1 200 OK 
Server: Microsoft-IIS/5.0 
Content-Location: http://iis.example.com/Default.htm 
Date: Fri, 01 Jan 1999 20:13: 52 GMT 
Content-Type: text/HTML 
Accept-Ranges: bytes 
Last-Modified: Fri, 01 Jan 1999 20:13: 52 GMT 
ETag: W/e0d362a4c335be1: ae1 
Content-Length: 133 

Response from Netscape Enterprise 4.1

$ nc netscape.example.com 80 
HEAD / HTTP/1.0 

HTTP/1.1 200 OK 
Server: Netscape-Enterprise/4.1 
Date: Mon, 16 Jun 2003 06:01: 40 GMT 
Content-type: text/HTML 
Last-modified: Wed, 31 Jul 2002 15:37: 56 GMT 
Content-length: 57 
Accept-ranges: bytes 
Connection: close 

We can notice that the ordering of the Date field and the Server field differs between Apache, Netscape Enterprise and IIS.

Malformed requests test

Another useful test to execute involves sending malformed requests or requests of nonexistent pages to the server. We consider the following HTTP response:

Response from Apache 1.3.23

$ nc apache.example.com 80 
GET / HTTP/3.0 

HTTP/1.1 400 Bad Request 
Date: Sun, 15 Jun 2003 17:12: 37 GMT 
Server: Apache/1.3.23 
Connection: close 
Transfer: chunked 
Content-Type: text/HTML; charset=iso-8859-1 

Response from IIS 5.0

$ nc iis.example.com 80 
GET / HTTP/3.0 

HTTP/1.1 200 OK 
Server: Microsoft-IIS/5.0 
Content-Location: http://iis.example.com/Default.htm 
Date: Fri, 01 Jan 1999 20:14: 02 GMT 
Content-Type: text/HTML 
Accept-Ranges: bytes 
Last-Modified: Fri, 01 Jan 1999 20:14: 02 GMT 
ETag: W/e0d362a4c335be1: ae1 
Content-Length: 133 

Response from Netscape Enterprise 4.1

$ nc netscape.example.com 80 
GET / HTTP/3.0 

HTTP/1.1 505 HTTP Version Not Supported 
Server: Netscape-Enterprise/4.1 
Date: Mon, 16 Jun 2003 06:04: 04 GMT 
Content-length: 140 
Content-type: text/HTML 
Connection: close 

We notice that every server answers in a different way. The answer also differs in the version of the server. An analogous issue comes if we create requests with a non-existant protocol. Consider the following responses:

Response from Apache 1.3.23

$ nc apache.example.com 80 
GET / JUNK/1.0 

HTTP/1.1 200 OK 
Date: Sun, 15 Jun 2003 17:17: 47 GMT 
Server: Apache/1.3.23 
Last-Modified: Thu, 27 Feb 2003 03:48: 19 GMT 
ETag: 32417-c4-3e5d8a83 
Accept-Ranges: bytes 
Content-Length: 196 
Connection: close 
Content-Type: text/HTML 

Response from IIS 5.0

$ nc iis.example.com 80 
GET / JUNK/1.0 

HTTP/1.1 400 Bad Request 
Server: Microsoft-IIS/5.0 
Date: Fri, 01 Jan 1999 20:14: 34 GMT 
Content-Type: text/HTML 
Content-Length: 87 

Response from Netscape Enterprise 4.1

$ nc netscape.example.com 80 
GET / JUNK/1.0 

<HTML><HEAD><TITLE>Bad request</TITLE></HEAD> 
<BODY><H1>Bad request</H1> 
Your browser sent to query this server could not understand. 
</BODY></HTML>

Automated Testing

As can be appreciated from above, a number of tests are needed in order to accurately determine the type and version of web server in use. There are tools available which can automate these tests. One such example is "httprint" which compares the web server response with a signature dictionary and then makes a judgement with associated confidence rating.
An example of such tool is shown below:

Bad request

</BODY></HTML>

Automated Testing

As can be appreciated from above, a number of tests are needed in order to accurately determine the type and version of web server in use. There are tools available which can automate these tests. One such example is "httprint" which compares the web server response with a signature dictionary and then makes a judgement with associated confidence rating.
An example of such tool is shown below: