Blind XPath Injection
Last revision (mm/dd/yy): 05/30/2013
XPath is a type of query language that describes how to locate specific elements (including attributes, processing instructions, etc.) in an XML document. Since it is a query language, XPath is somewhat similar to Structured Query Language (SQL), however, XPath is different in that it can be used to reference almost any part of an XML document without access control restrictions. In SQL, a "user" (which is a term undefined in the XPath/XML context) may be restricted to certain databases, tables, columns, or queries. Using an XPATH Injection attack, an attacker is able to modify the XPATH query to perform an action of his choosing.
Blind XPath Injection attacks can be used to extract data from an application that embeds user supplied data in an unsafe way. When input is not properly sanitized, an attacker can supply valid XPath code that is executed. This type of attack is used in situations where the attacker has no knowledge about the structure of the XML document, or perhaps error message are suppressed, and is only able to pull once piece of information at a time by asking true/false questions(booleanized queries), much like Blind SQL Injection.
For more information, please see the article on regular XPATH Injection.
The attacker may mount a successful attack using two methods: Boolenization and XML Crawling. By adding to the XPath syntax, the attacker uses additional expressions (replacing what the attacker entered in the place of the injection).
Using the "Boolenization" method the attacker may find out if the given XPath expression is True or False. Let's assume that the aim of the attacker is to log in to an account in a web application. A Successful log in would return "True" and failed log in attempt would return "False". Only a small portion of the information is targeted via the analyzed character or number. When the attacker focuses on a string he may reveal it in its entirety by checking every single character within the class/range of characters this string belongs to.
Using a string-length(S) function, where S is a string, the attacker may find out the length of this string. With the appropriate number of substring(S,N,1) function iterations, where S is a previously mentioned string, N is a start character, and "1" is a next character counting from N character, the attacker is able to enumerate the whole string.
<?xml version="1.0" encoding="UTF-8"?> <data> <user> <login>admin</login> <password>test</password> <realname>SuperUser</realname> </user> <user> <login>rezos</login> <password>rezos123</password> <realname>Simple User</realname> </user> </data>
- string.stringlength(//user[position()=1]/child::node()[position()=2]) returns the length of the second string of the first user (8),
- substring((//user[position()=1]/child::node()[position()=2),1,1) returns the first character of this user ('r').
To get to know the XML document structure the attacker may use:
This will return the number of nodes (in this case 2).
Using this query the attacker will find out if the second string (password) of the first node (user 'admin') consists of 6 characters.
- substring(string, number, number)
This query will confirm (True) or deny (False) that the first character of the user ('admin') password is an "a" character.
If the log in form would look like this:
String FindUser; FindUser = "//user[login/text()='" + Request("Username") + "' And password/text()='" + Request("Password") + "']";
then the attacker should inject the following code:
Username: ' or substring((//user[position()=1]/child::node()[position()=2]),1,1)="a" or ''='
The XPath syntax may remind you of common SQL Injection attacks but the attacker must consider that this language disallows commenting out the rest of expresssion. To omit this limitation the attacker should use OR expressions to void all expressions, which may disrupt the attack.
Because of Boolenization the number of queries, even within a small XML document, may be very high (thousands, houndred of thousands and more). That is why this attack is not conducted manually. Knowing a few basic XPath functions, the attacker is able to write an application in a short time which will rebuild the structure of the document and will fill it with data by itself.
Related Threat Agents
- http://dl.packetstormsecurity.net/papers/bypass/Blind_XPath_Injection_20040518.pdf - by Amit Klein (much more detailes, in my opinion the best source about Blind XPath Injection).