Spring View Manipulation
Views overview
Consider a simple Spring application that uses Thymeleaf as its templating engine:
The index
method will be called for every GET
request for the root url /
. It has no parameters and returns the static string welcome
. Spring interprets welcome
as the name of the View
and tries to find the resources/templates/welcome.html
file located in the application resources. If Spring finds it, renders a view from the template file and returns to the user.
Untrusted Thymeleaf view name
If Thymeleaf view engine is used (the most popular for Spring), the template might look like this one:
Thymeleaf engine supports file layouts, that allows you to specify a fragment in the template by using <div th:fragment="main">
and then request only this fragment from the view:
Thymeleaf is intelligent enough to return only the main div
from the welcome
view, but not the whole document.
Before loading the template from the filesystem, Spring ThymeleafView class parses the template name as an expression:
As a result, if template name or fragment is concatenated with untrusted data, it can lead to expression language injection and hence RCE.
For instance, the following methods are vulnerable to expression language injection:
The following request creates the executed
file on the server:
This exploit uses expression preprocessing. In order for the expression to be executed by the Thymeleaf, no matter what prefixes or suffixes are, it is necessary to surround it with __${
and }__::.x
.
Untrusted implicit view name
Controllers do not always return strings that explicitly tell Spring what view name to use. As described in the documentation, for some return types such as void
, java.util.Map
or org.springframework.ui.Model
, the view name is implicitly determined through a RequestToViewNameTranslator
.
This means that at first glance such a controller may seem completely innocent, it does almost nothing, but since Spring does not know which view name to use, it takes it from the request URI.
Specifically, DefaultRequestToViewNameTranslator
does the following:
So, it becomes vulnerable because the user controlled data (URI) comes in directly to view name and is resolved as an expression:
References
Last updated