|
|
|
|
| |
Credit:
The information has been provided by Core Security Technologies Advisories.
The original article can be found at: http://www.coresecurity.com/index.php5?module=ContentMod&action=item&id=1924
|
| |
Vulnerable Systems:
* AIM version 6.1 (6.1.41.2)
* AIM version 6.2 (6.2.32.1)
* AIM version Pro
* AIM version Lite
Immune Systems:
* AIM version 6.5 (6.5.3.12) - http://beta.aol.com/projects.php?project=aim6
* AIM version Express - http://www.aim.com/aimexpress.adp
* Classic AIM version 5.9 - http://www.aim.com/get_aim/win/other_win.adp
Vendor Information, Solutions and Workarounds:
AOL has become aware of security vulnerabilities in several AIM instant messaging clients. Successful exploitation of these vulnerabilities could allow an attacker to execute arbitrary commands on a user's workstation. AOL has deployed host side filtering on the AIM servers to block this potentially malicious content from being sent to AIM clients.
Solutions:
1. Users of AIM can upgrade to the latest version of the AIM beta client at beta.aol.com.
Other workarounds (un-official)
Workaround #1: Users running AIM on Microsoft Windows XP SP2 or Windows Server 2003 SP1 may implement Microsoft's "Internet Explorer Local Machine Zone Lockdown" recommendations to mitigate risk. This will not fix the reported bugs but will reduce the risk of exploitation significantly. To enable Local Machine Zone Lockdown for your AIM client, go to the following registry key:
HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft \Internet Explorer\Main\FeatureControl\FEATURE_LocalMachine_Lockdown
Add a REG_DWORD value to this key named as the AIM client application (for example, aim.exe) and set it to 1. Any other setting for this value will disable Local Machine Zone Lockdown for the application.
For further details about how to configure this feature read Microsoft's Internet Explorer Local Machine Zone Lockdown recommendation at: http://technet.microsoft.com/enus/library/bb457150.aspx#EHAA
Technical Description / Proof of Concept Code:
The standard protocol that AIM clients use to communicate is called OSCAR (Open System for CommunicAtion in Realtime), which is a closed protocol also used by AOL's secondary Instant Messaging client, ICQ (I Seek You). On top of the OSCAR protocol, AIM clients have implemented support for enhanced message types that use features provided by the HTML (Hyper Text Markup Language) in order to, for example, provide AIM users with the possibility of exchanging text messages with specific font formats or colors. AIM 6.1, AIM 6.2 (beta), AIM Pro and AIM Lite have embedded an Internet Explorer server control in the message display window in order to facilitate the parsing and displaying of HTML controls. It is a common practice for Windows applications to reuse Microsoft Internet Explorer's HTML parsing objects included in the mshtml.dll library instead of using a homebrew HTML parser.
Several programming frameworks, including MFC (Microsoft Foundation Classes), provide practical ways of embedding such controls through classes like CHtmlEditView or CHtmlEditDoc. Some of the advantages of using MSHTML are that it provides a particular, feature-rich and somewhat complete support for DHTML and also that it is easier to host Microsoft ActiveX Controls. However, in the context of this advisory, such advantages may end up becoming security problems due to design flaws and implementation bugs. There are two particular characteristics in the implementation of the described functionality that turn AIM's highly flexible message-content features into high-risk attack vectors for its users.
First, the vulnerable IM clients do most of the sanitizing/filtering and encoding of HTML content on outbound messages, thus a malicious attacker with the ability to bypass outbound HTML filtering can send any type of HTML content to other IM clients.
A handful of publicly available and well-known IM clients permit to send un-sanitized data to any other client that supports the same communications protocol including the vulnerable AIM 6.1, AIM 6.2, AIM Pro and AIM Lite clients.
Second, although there are some defensive mechanisms implemented in the vulnerable clients these are insufficient to properly handle messages with potentially malicious content. Input validation of inbound messages appears to be taking place but can be easily circumvented by an attacker.
As a result, the entire attack surface of MSHTML is exposed to remote IM peers. By having a way of sending data straight to the MSHTML library, attackers could abuse such high-risk attack vector to:
- Execute arbitrary shell commands in the victim's workstation.
- Direct the embedded IE to perform arbitrary HTTP requests (CSRF)
- Include HTML controls (links, images, forms ) in IM text messages in order to trick users into revealing sensitive information or performing harmful actions against their accounts/workstation/etc.
- Run JavaScript code within IE to enhance the attacks mentioned above.
- Instantiate ActiveX controls, which attackers could use to target vulnerabilities in the ActiveX objects themselves or use their functionality to, for example, read arbitrary files from the victim's file system or even execute arbitrary shell commands in the victim's workstation.
- Directly attack vulnerable versions of Internet Explorer in user workstations. This is a typical client-side attack scenario and could lead to the remote execution of arbitrary code in the victim's workstation. In this scenario "one-click" IE bugs (exploitation requires user assistance) become "zero-click" bugs (exploitation does not require user interaction).
AOL's "Classic AIM 5.9" is an official alternative client for nostalgic users and is not vulnerable due to the fact that instead of using MSHTML to render HTML it appears to include limited rendering functionality either provided by a third party library or homebrew code. Although there is no guarantee that its implementation lacks vulnerabilities, in our tests it did prevent the attack vectors described in this advisory. So is the case for AOL's AOL 6.5.3.12 which although it is embedding an Internet Explorer server control in the message window, could not be exploited during our tests.
AOL's online AIM Express web client which is written in ASP.NET and also appears to be taking the necessary defensive measures required to prevent any of these problems from taking place.
Proof of concept snippets:
The following examples provide code snippets that should serve as proof-of-concept code to demonstrate some of the problems that arise from the issue reported. The snippets have been arranged according to their risk level, in increasing order (lower risk first), with the intention of making this process more self-explanatory. In order for these snippets to work, they must be sent within the contents of a standard instant message, but using a client that will not encode message contents on output:
Using HTML controls in order to trick victims into revealing sensitive information or do harmful actions against their accounts/workstations or to force outbound HTTP requests (CSRF).
The following proof-of-concept code was successfully tested on AIM Pro. Other vulnerable clients need some tweaking in order to get it to work. The code uses <hr>, <h2>, <form> and <input> tags to trick the victim into believing that an AIM disconnection took place. It then requests the victim to type in the user credentials and to press the Reconnect (submit) button in order to send the credentials to the attacker. In this case, the form is submitted to "http://localhost"
--- begin code ---
<hr><h2>SERVICE UNAVAILABLE</h2><hr>Your connection to AIM has been
interrupted. Please type in your credentials in order to reconnect. Thank
you.<br><form method=GET action=http://localhost>Login: <input type=text
name=login><br>Password: <input type=text name=password><br><input
type=submit value=Reconnect></form><br><hr>
--- end code ---
This is simply one example of exploitation using only embedded HTML. There are plenty of HTML controls that could be used in similar exploitation scenarios.
Using scripting languages to enhance an attack
As mentioned in the beginning of the technical details section, we identified among all the vulnerable clients what appeared to be an existing defensive measure (or a functional characteristic with a side-effect) meant to prevent attackers from inserting malicious JavaScript statements within message contents. When sending JavaScript statements inside <script> tags within a message, the script tags appear to be filtered by the receiving client upon display (therefore not executing the JavaScript statements). However, this was easily circumvented by embedding JavaScript statements inside <img> tags, as in the following example:
The following code does not work:
<script>alert( I have executed javascript in your computer )</script>
The following code *does* work:
<img src="javascript:alert( I have executed javascript in your computer )">
Note that even though the different proof-of-concept snippets provided in this advisory use <img> tags to execute JavaScript, the problem must not be thought of as circumscribed to message contents with embedded <img> tags only. JavaScript statements may also be executed through the use of other HTML controls and some of the attack vectors that we mention do not even rely on JavaScript for successful exploitation.
The following proof-of-concept code will display a prompt box to the victim, requesting to type in the victim's AIM credentials. It will look authentic due to the fact that the message box is not part of the text message window:
-- begin code --
<img src='javascript:var passwd=window.prompt("You have been
disconnected from the network.\nPlease re-enter your password to
reconnect.", "");alert( You have disclosed your password: +passwd);'>
-- end code --
Once the victim types-in her/his password, an alert message box will be displayed showing the entered password. Attackers could easily retrieve passwords and other security-sensitive data by using the same techniques used to exploit Cross Site Scripting vulnerabilities to steal browser cookies.
Instantiating ActiveX controls and taking over the victim's workstation
Another way of enhancing an attack could rely on using ActiveX controls installed on the target system. For that, the attacker needs the ability to instantiate arbitrary ActiveX controls using an IM message content constructed to accomplish such purpose. We successfully used this attack vector in AIM 6.1, AIM 6.2beta and AIM Lite in order to get immediate and instant access to the victim's workstation. This attack vector increases considerably the severity of the problems found, turning the affected clients into a doorway to the user's computer and ultimately providing attackers with ways of executing arbitrary commands.
Apparently, AIM Pro is the only client that runs the Internet Explorer server control in a "protected" security zone, where the victim is prompted with the typical message box that says:
"An ActiveX control on this page might be unsafe to interact with other parts of the page. Do you want to allow this interaction?"
The choice of the user will affect the entire instance of the application and be applied to any other existing/future message windows (as well as potentially any other locations where the Internet Explorer server control is used.)
Attackers could use JavaScript to instantiate ActiveX controls in order to either exploit client-side vulnerabilities of the ActiveX objects themselves or to use ActiveX functionality as an aid to exploit other bugs. As the following proof-of-concept snippet shows, an attacker can successfully instantiate the "Shell.Application" object that is included in the Microsoft Windows OS installation and use it to execute any arbitrary command on the victim's workstation.
As previously mentioned, in three of the four affected versions of the product, the attack is straight forward and no interaction with the victim is required. Such clients appear to be running Internet Explorer control in the Local Machine Security Zone.
-- begin code --
<img src="javascript:var oShell = new
ActiveXObject('Shell.Application');oShell.ShellExecute('cmd.exe', '/c
pause');">
-- end code --
The proof-of-concept code from above will run an instance of Microsoft Windows command line tool, executing the pause command. Upon receipt, it will instantly show a blank command window in the victim's workstation. An attacker may easily abuse this issue to gain complete control over the victim's machine with the privileges of the user running AIM.
Attacking vulnerable versions of Internet Explorer controls
This scenario is just a clear-cut client-side attack vector and would rely on any unpatched security vulnerability in Internet Explorer or the ActiveX controls it hosts. An embedded HTML URI can direct the IE server control to automatically visit a previously setup site that delivers malicious content to exploit known Internet Explorer vulnerabilities. In this case, the AIM clients identified as vulnerable in this advisory play
the role of exposing their users to attacks without requiring user intervention to allow/disallow access to the rogue website. This attack scenario is functionally equivalent to a user of Internet Explorer clicking on a URL and visiting a malicious site.
Additional Information and resources
AOL's AIM clients
* AIM Pro: http://aimpro.premiumservices.aol.com/
* AIM 6.2: http://beta.aol.com/projects.php?project=aim6
* AIM 6.1: http://www.aim.com/get_aim/win/latest_win.adp
* AIM Lite: http://x.aim.com/laim/
* AIM 5.9: http://www.aim.com/get_aim/win/other_win.adp
* AIM 6.5: http://beta.aol.com/projects.php?project=aim6
Embedding IE
* Introduction to ActiveX Controls: http://msdn2.microsoft.com/en-us/library/aa751972.aspx
Reusing MSHTML:
* http://msdn2.microsoft.com/en-us/library/bb508516.aspx
Internet Explorer Local Machine Zone Lockdown
* http://technet.microsoft.com/en-us/library/bb457150.aspx#EHAA
* http://technet2.microsoft.com/windowsserver/en/library/aebcfc94-25d5-4f41-93cc-7fb6e031de401033.mspx?mfr=true
About URL Security Zones
* http://msdn2.microsoft.com/en-us/library/ms537183.aspx
References
[1] AIM Arbitrary HTML Display in Notification Window shell at dotshell dot net http://www.securityfocus.com/archive/1/479199
[2] AIM Local File Display in Notification Window shell at dotshell dot net http://www.securityfocus.com/archive/1/479435
[3] http://nvd.nist.gov/nvd.cfm?cvename=CVE-2007-4901
[4] http://www.securityfocus.com/bid/25659
[5] http://secunia.com/advisories/26786/
[6] "AIM 6.1 problems" thread on AOL's AIM Support & more technical forum http://messageboards.aol.com/aol/en_us/articles.php?boardId=565563&articleId=16537
[7] "IM problems" thread in AOL's AIM 6 Technical Issues forum http://messageboards.aol.com/aol/en_us/articles.php?boardId=565563&articleId=16537
[8] "Copyright and Confidentiality notice?" thread on AOL's AIM 6 Technical Issues forum http://messageboards.aol.com/aol/en_us/articles.php?boardId=567774&articleId=2400
|
|
|
|
|
