XPATH Injection

Similar to SQL Injection, XPath Injection attacks occur when a web site uses user-supplied information to construct an XPath query for XML data. By sending intentionally malformed information into the web site, an attacker can find out how the XML data is structured, or access data that he may not normally have access to. He may even be able to elevate his privileges on the web site if the XML data is being used for authentication (such as an XML based user file).
Querying XML is done with XPath, a type of simple descriptive statement that allows the XML query to locate a piece of information. Like SQL, you can specify certain attributes to find, and patterns to match. When using XML for a web site it is common to accept some form of input on the query string to identify the content to locate and display on the page. This input must be sanitized to verify that it doesn't mess up the XPath query and return the wrong data.
XPath is a standard language; its notation/syntax is always implementation independent, which means the attack may be automated. There are no different dialects as it takes place in requests to the SQL databases.

Because there is no level access control it's possible to get the entire document. We won't encounter any limitations as we may know from SQL injection attacks.

Example

Suppose we have a user authentication system on a web page that used a data file of this sort to login users. Once a username and password have been supplied the software might use XPath to look up the user:

VB:
Dim FindUserXPath as String
FindUserXPath = "//Employee[UserName/text()='" & Request("Username") & "' And 
        Password/text()='" & Request("Password") & "']"

C#:
String FindUserXPath;
FindUserXPath = "//Employee[UserName/text()='" + Request("Username") + "' And 
        Password/text()='" + Request("Password") + "']";

With a normal username and password this XPath would work, but an attacker may send a bad username and password and get an XML node selected without knowing the username or password, like this:

Username: blah' or 1=1 or 'a'='a
Password: blah

FindUserXPath becomes //Employee[UserName/text()='blah' or 1=1 or 
        'a'='a' And Password/text()='blah']

Logically this is equivalent to:
        //Employee[(UserName/text()='blah' or 1=1) or 
        ('a'='a' And Password/text()='blah')]

In this case, only the first part of the XPath needs to be true. The password part becomes irrelevant, and the UserName part will match ALL employees because of the "1=1" part.

XPath Injection Defenses

Just like the techniques to avoid SQL injection, you need to use a parameterized XPath interface if one is available, or escape the user input to make it safe to include in a dynamically constructed query. If you are using quotes to terminate untrusted input in a dynamically constructed XPath query, then you need to escape that quote in the untrusted input to ensure the untrusted data can't try to break out of that quoted context. In the following example, single quotes (') are used to terminate the Username and Password parameters. So, we need to replace any ' characters in this input with the XML encoded version of that character, which is "'".

VB:
Dim FindUserXPath as String
FindUserXPath = "//Employee[UserName/text()='" & Request("Username").Replace("'", "'") & "' And 
        Password/text()='" & Request("Password").Replace("'", "'") & "']"

C#:
String FindUserXPath;
FindUserXPath = "//Employee[UserName/text()='" + Request("Username").Replace("'", "'") + "' And 
        Password/text()='" + Request("Password").Replace("'", "'") + "']";

Another better mitigation option is to use a precompiled Xpath query. Precompiled XPath queries are already preset before the program executes, rather than created on the fly after the user's input has been added to the string. This is a better route because you don't have to worry about missing a character that should have been escaped.