Sunday, January 23, 2011

Enhanced ASP.NET input sanitation with AntiXSS

Need for input sanitation

The applications you build for the internet can have quite a big audience. It is possible for the whole connected world to visit your website and use the app you built on it. But with these visitors, you will attract dark forces who can misuse your application for their own agenda. It is therefore needed to harden your application, and be sure no-one can send email, inject custom HTML and javascript and hijack user information. It is needed you built secure code and protect yourself.

The main characteristic of an application is that it needs input data from the end user. With this input data, something is done. If this data is sent back to the user, is can be displayed on the screen. In it most simple form, this looks like:


<div>
<asp:TextBox ID="inputBox" runat="server" TextMode="MultiLine" />
<br />
<asp:Button ID="submitButton" runat="server" Text="Submit"
onclick="submitButton_Click" />
<br />
<asp:Label runat="server" EnableViewState="True" AssociatedControlID="resultLiteral" Text="Html output:" />
<asp:Literal ID="resultLiteral" runat="server"></asp:Literal>
</div>



With the following button_click code:


protected void submitButton_Click(object sender, EventArgs e)
{
string input = inputBox.Text;
string output = input;
resultLiteral.Text = output;
}


With this example code, it is possible to inject javascript into the response. Try pasting <script language="javascript">alert('Hello World!');</script> in the input box and see.

Out of the box ASP.NET functionality

With the standard ASP.NET setup, the last script example didn’t worked, but threw an exception with the message ‘A potentially dangerous Request.Form value was detected from the client (inputBox="<script language="ja...").’. This is because out of the box, ASP.NET will scan each input parameter to dangerous characters. But don’t let this filter mislead you. The default characters will be the less than character (<) and some ampersand-hash encoded characters. There are also scenario’s you will be switching off the request validation. Simply because some business scenario (maybe you want some html from your user).

You can switch your validation off for all your pages in ASP.NET in the web.config, or specify ValidateRequest="false" in the page directive for the specific page you want to disable this check (note that in ASP.NET 4, there are some additional steps needed). But don’t rely on request validation to heavily, because there are situations in can’t detect malicious input.

Shortcomings of Request Validation

Think about the following situation. You want the user of your site to specify the background color of an element on your page, and you use the following submit event:

protected void submitButton_Click(object sender, EventArgs e)
{
string color = inputBox.Text;
string output = "<div style=\"width:200px;height:200px;background-color:" + color + ";\">Colored Box</div>";
resultLiteral.Text = output;
}


Because the input is directly pasted into an attribute of the div tag, you don’t need HTML directly to inject javascript. The following line will pass through input validation, and inject the page with some potentially evil javascript:
red;" onclick="alert('Hello World!');" "
/



Filtering input

The first thing to prevent the injection of those unwanted code, is to filter the user input if used for generating output. Standard built-in in ASP.NET is the HttpUtility class with some Encoding methods. They can help with the above two attacks, but it is still possible to inject URL backgrounds in the second example. Because these filters are limited in usage, Microsoft started a new library quite some time ago.

AntiXSS library

The two main features in the AntiXSS library are the Encoder and Sanitizer classes. Like the HttpUtility, the encoder class has several specific Encoding methods. These methods are a lot more specific than HttpUtility though. In the second example, it is possible filter the color input with a call to Encoder.CssEncode. The output will be CSS encoded into ‘red\00003B\000022\000020onclick\00003D\000022alert\000028\000027Hello\000020World\000021\000027\000029\00003B\000022\000020\000022’.

The Sanitizer class is even more powerful. It can strip certain HTML tags out of the input. It can transform <script language="javascript">alert('Hello World!');</script><strong>bla</strong><img src="bla.gif" /><ul><li>one<li>Two<li>Three</ul> into <strong>bla</strong><img src="bla.gif"> <ul> <li>one</li><li>Two</li><li>Three</li></ul>


Install AntiXSS with NuGet

AntiXSS used to be a very hidden gem somewhere deep on Microsoft Download or CodePlex. But with the new NuGet, it is on page one of packages. First you need to install NuGet into your Visual Studio. After that, right click your project and choose ‘Add Library Package Reference’.



imageimage


As of today, AntiXSS is the eighth package in the list. If you click install, all assemblies for AntiXSS will be automatically downloaded and added to your solution. So it won’t be very hard anymore to write really secure code!

1 comment:

abhijit kharde said...

Thanks for the nice post. However currently I am facing issue while installing NuGet package as mentioned in this article like Id/Author cannot be null. Any idea how to resolve this?

Share