Can't create CSRF token with Spring Security - spring-mvc

I am using Spring Security 3.2.3 in my Spring MVC application and getting some unexpected behavior.
According to the documentation here, it should be possible to use ${_csrf.token} in the meta tags of my html:
<meta name="_csrf" content="${_csrf.token}" />
<!-- default header name is X-CSRF-TOKEN -->
<meta name="_csrf_header" content="${_csrf.headerName}" />
From where I extract the value of "content" using JQuery and place it into the Request Header using AJAX.
For some reason though, Spring Security doesn't "convert" this into an actual token, it just gets sent into the header as a literal string "${_csrf.token}".
Trying the alternate route of using ${_csrf.token} in a hidden input according to the documentation, I then tried to check what the token evaluates to by checking the input's value, but it's still just plain text "${_csrf.token}".
Since it seems that Spring Security isn't in effect, am I missing some kind of configuration? I am currently using a barebones Spring Security Java configuration (as opposed to xml) as shown here:
import org.springframework.context.annotation.*;
import org.springframework.security.config.annotation.web.builders.HttpSecurity;
import org.springframework.security.config.annotation.web.configuration.*;
#Configuration
#EnableWebSecurity
public class WebSecurityConfig extends WebSecurityConfigurerAdapter {
#Override
protected void configure(HttpSecurity http) throws Exception {
http
.csrf();
}
}
I know configure is getting called since I put a debug statement in it, so I assume that CSRF protection is indeed enabled since it should be by default.
I realize that the syntax "${}" is JSP Expression Language, and I am currently successfully using it to evaluate the context into an object with Thymeleaf, for example:
th:object="${context}"
So I tried adding "th:" in front of the meta tag's "content" like so:
<meta name="_csrf" th:content="${_csrf.token}"/>
But it results in an exception that this cannot be evaluated:
Exception evaluating SpringEL expression: "_csrf.token"
I think the key here may be figuring out how to get the expression to evaluate properly in my view.

I finally solved this problem, but it basically required rewriting Spring Security. Here it is in all its glory.
First, I followed the suggestions in Eyal Lupu's great blog post here, but I had to tweak it to my situation because of my AJAX requirement.
As for the Thymeleaf situation, the key tidbit is hidden away in the archives of the Thymeleaf forums - Infamous Issue 7.
https://github.com/thymeleaf/thymeleaf-spring/issues/7#issuecomment-27643488
The last comment by the creator of Thymeleaf himself says that:
th:action ... detects when this attribute is being applied on a
tag --which should be the only place, anyway--, and in such case
calls RequestDataValueProcessor.getExtraHiddenFields(... ) and adds the
returned hidden fields just before the closing tag.
That was the key phrase I needed to get the token to work. Unfortunately it's completely not obvious why th:action would also kick off getExtraHiddenFields, but at any rate it does, and that's what matters.
So for anyone struggling with Thymeleaf + Spring Security CSRF + AJAX POST, here are my steps (this is paring it down quite a bit but these are the high-level concepts to solve it):
Implement the Spring interface RequestDataValueProcessor and register it in Spring Security's XML config so you can override the method getExtraHiddenFields, which allows you to insert a hidden input field into the HTML (with the token of course). The token itself is generated with a Java.Util UUID.
With JQuery, read the value from that hidden field and set the Request Header's "X-CSRF-Token" attribute so that it gets sent over HTTP. It's not possible to simply leave the token in the hidden input field because we are not doing a form Submit, instead we use AJAX POST to call methods on the server side.
Extend Spring's HandlerInterceptorAdapter and register it as an interceptor so that every time a POST method is done, the "preHandle" method on the server side is called so it can compare the request token (extracted from the HTTP header in the previous step) to the session's token (should be the same!). After it does this check, it can either allow the request to go through or return an error.

I started with the same source article as you, I think, and the same "you should be able to" add answers as you did. I fought it a different way. I made Thymeleaf give me the answer I wanted.
<meta name="_csrf" th:content="${_csrf.token}"/>
<!-- default header name is X-CSRF-TOKEN -->
<meta name="_csrf_header" th:content="${_csrf.headerName}"/>
Thymeleaf put the attribute "content" with the requested Spring EL contents. I then used the provided JavaScript/JQuery to extract the info from the meta tags straight into the CSRF header.

Before adding the thymeleaf-extras-springsecurity namespace and its dependency into my project, I had similar problems. I never did get the meta tags to work, even with thymeleaf-extras-springsecurity. But I did successfully retrieve Spring Security's csrf token using the hidden input. I have instructions below that work for me:
In the html tag, add:
xmlns:sec="http://www.thymeleaf.org/thymeleaf-extras-springsecurity4"
In your pom.xml (if you're using Maven) you'll need to add the dependency: thymeleaf-extras-springsecurity4.
Then add the hidden input inside your page's body to retrieve the csrf token.
<input type="hidden" id= "csrf-token" th:name="${_csrf.parameterName}" th:content="${_csrf.token}" />
and then use that within your javascript/jquery as follows:
function f1() {
var token1 = $('input#csrf-token').attr("content");
...
$.ajax({
...
type: "POST",
beforeSend: function (request)
{
request.setRequestHeader("X-CSRF-TOKEN", token1);
},
...
This all assumes that you have spring security enabled, and that you have NOT turned off csrf protection.

You have incorrect configuration for springSecurityFilterChain in your web.xml. Correct definition is:
<filter>
<filter-name>springSecurityFilterChain</filter-name>
<filter-class>org.springframework.web.filter.DelegatingFilterProxy</filter-class>
</filter>
Spring Security uses set of servlet filters to provide the functionality it is offering (including CSRF protection). These filters are defined as Spring beans (i.e. they are instantiated and managed by Spring application context). DelegatingFilterProxy is a special type of servlet filter, which finds root application context on the registered servlet context and delegates every call to the same named bean.

Your issue is a different one, just stumbled across this one as well and it took me several hours to figure out the cause. The cause for the issue described by you is that you did not enable csrf support within your spring-security.xml
This little snippet needs to go into your security-config.xml:
<!-- Static resources such as CSS and JS files are ignored by Spring Security -->
<security:http pattern="/static/**" security="none" />
<security:http use-expressions="true">
<!-- Enables Spring Security CSRF protection -->
<security:csrf/>
<!-- Configures the form login -->
<security:form-login
login-page="/login"
login-processing-url="/login/authenticate"
authentication-failure-url="/login?error=bad_credentials"
username-parameter="username"
password-parameter="password"/>
<!-- Configures the logout function -->
<security:logout
logout-url="/logout"
logout-success-url="/login"
delete-cookies="JESSIONID"/>
<!-- Anyone can access these urls -->
<security:intercept-url pattern="/auth/**" access="permitAll"/>
<security:intercept-url pattern="/login" access="permitAll"/>
<security:intercept-url pattern="/signin/**" access="permitAll"/>
<security:intercept-url pattern="/signup/**" access="permitAll"/>
<security:intercept-url pattern="/user/register/**" access="permitAll"/>
<!-- The rest of our application is protected. -->
<security:intercept-url pattern="/**" access="hasRole('ROLE_USER')"/>
<!-- Adds social authentication filter to the Spring Security filter chain. -->
<security:custom-filter ref="socialAuthenticationFilter" before="PRE_AUTH_FILTER" />
</security:http>
....
...
..
.
Save time by configuring this correctly...
Cheerio,
Flo!

In case you don't need to use Thymeleaf, I'd suggest the following:
Add this to the top of your page:
<%# taglib prefix="c" uri="http://java.sun.com/jsp/jstl/core"%>
Add this to your login form:
<input type="hidden" name="${_csrf.parameterName}" value="${_csrf.token}" />
Add these dependencies to your pom.xml:
<dependency>
<groupId>org.apache.tomcat.embed</groupId>
<artifactId>tomcat-embed-jasper</artifactId>
<scope>provided</scope>
</dependency>
<dependency>
<groupId>javax.servlet</groupId>
<artifactId>jstl</artifactId>
</dependency>
After struggling a lot, that worked for me.

Related

Denying Access To Certain Urls

I would like to block the access to certain URLs in my spring web application. I did some research and came across "headers". I don't know how to use them though. Any help will be highly appreciated. Thanks in advance.
One of the most simple way to do this, which I can think of instantly, is to have a pre-defined list of blocked urls and check the request URL against this list in the overridden requiresAuthentication() method of your implementation of AbstractAuthenticationProcessingFilter.
#Override
protected boolean requiresAuthentication(final HttpServletRequest request, final HttpServletResponse response) {
List<String> blockedUrls = new Arraylist<>();
list.add("/foo/");
if(blockedUrls.contains(request.getRequestURI())) {
// go to blocked error page
} else {
// process the request normally
}
}
Shishir
How is your spring security configured? can you show us the code/xml?
Assuming you have configured spring-security with your webapp then it is easy to restrict URLs (including role based restrictions).
<http auto-config="true" use-expressions="true">
<intercept-url pattern="/restricted/**" access="isAuthenticated()" />
<intercept-url pattern="/**" access="permitAll" />
<form-login login-processing-url="/loginProcess"
login-page="/"
authentication-failure-url="/?loginFailure=true"/>
<logout logout-url="/static/j_spring_security_logout"/>
</http>
The intercept-url elements define URL patterns and whether anyone/everyone can access that pattern. Spring will match from the top, so the most specific authentication should be at the top of the list. Also, in the above example, as the final entry matches all URLs, this means anything not specified in other URL patterns will be open.
Taken from this article - it also shows how to achieve the same using code config rather than in xml: http://automateddeveloper.blogspot.co.uk/2014/02/spring-4-xml-to-annotation-configuration.html

JBoss Seam only allow Parameters via POST

I'm using JBoss 4.2 with Java Seam and in my application there is an login-form implemented.
Now if I submit the form in the GUI the Parameters will be sent via HTTP-POST, but if I try to send the Data via HTTP-GET it will be accepted too.
That should be prevented, but I didn't find out how.
Could I set the Method in the pages.xml or do I need to find out programmatically if the Request is Post or Get.
Are there other ways to do that or how should I do it?
Thanks!
Add the followin Statement in the web.xml:
<security-constraint>
<web-resource-collection>
<web-resource-name>Post Pages</web-resource-name>
<description />
<url-pattern>/*.xhtml</url-pattern>
<http-method>POST</http-method>
</web-resource-collection>
<user-data-constraint>
<transport-guarantee>NONE</transport-guarantee>
</user-data-constraint>
</security-constraint>
This will only allow POST Request in the defined url.patterns

SpringSecurity multiple namespaces and secured-annotations. Great confusion

I'm making a Spring MVC web-app with some RESTfull resources as an API.
I need the RESTfull part to have some custom filters as I do not want any redirection and I want any exception to be translated with the corresponding HTTP error code and a basic JSON description.
On the other hand, the rest of the website have to be more common and redirect people when they are not logged in etc.
One more thing, I wish to use the #Secured annotations and a post-authentication in some case.
How do I define the multiple http namespaces correctly (on Spring 3.1)?
Here is my erroneous configuration:
<global-method-security secured-annotations="enabled" />
<http pattern="/rest/**" authentication-manager-ref="authenticationManager" entry-point-ref="restAuthenticationEntryPoint">
<form-login login-page="/rest/login" login-processing-url="/rest/postlogin"
authentication-success-handler-ref="restAuthenticationSuccessHandler"
authentication-failure-handler-ref="restAuthenticationFailureHandler"
username-parameter="username" password-parameter="password" />
<logout logout-url="/rest/logout" invalidate-session="true" />
</http>
<http pattern="/**" authentication-manager-ref="authenticationManager">
<form-login login-page="/login" login-processing-url="/postlogin"
username-parameter="username" password-parameter="password" />
<logout />
</http>
The funny part is that this configuration works partially as I can login with /rest/login and I get the response from my custom success handler. I can also login from /login and I get the proper redirection to /. The logout are working both fine too.
Next, all the controllers beans have #Secured("ROLE_USER") in the secured methods. But all the secured methods don't ever get secured. Why is that so?
#Secured({"ROLE_USER"})
#RequestMapping(method = RequestMethod.GET, headers = { "Range" })
public #ResponseBody
HttpEntity<List<T>> list(#RequestHeader("Range") String range) {
I've read documentations everywhere and I'm more confused than ever.
Why are my methods not being secured?
Must the http namespace define an access so that the #Secured annotations work?
Are the http namespace overwriting my #Secured annotations? If it's so, how can I define multiple "login pages" with custom filters and being able to use annotations?
Here are some facts:
* I'm using Spring and SpringSecurity 3.1
* I have a custom AuthenticationManager to retrieve user details from hibernate daos.
* Some controllers are extending an abstract class where the #Secured annotations lies. But it still doesn't work for a simple controller.
* My controllers are discovered with a context:component-scan and a base-package.
* The security works fine with one http namespace.
please help, i'm getting mad with this!
Check out this answer about making sure the web context is visible to the global-method-security declaration and possibly using class proxying.
To answer your other questions, no the http namespace shouldn't affect the use of #Secured annotations, other than that the user is authenticated by the web part of the application and that information will be used by the method security interceptor when making an access decision. Unless you override it (using access-decision-manager-ref), method security will use a standard AccessDecisionManager which grants or denies access based on the roles a user has.

Exclude Spring MVC Controller from Spring Security

I have a web application secured with spring security (Spring 3.1.0). Now if a customer wants to register to my service, Spring Security say "No". This makes sense because the user is not yet authorized.
The controller, which gets the register data is a spring mvc controller. I need to exclude this from spring security I think.
I've excluded some urls so far like this:
<intercept-url pattern="/index.jsp" access="IS_AUTHENTICATED_ANONYMOUSLY" />
Is it possible, to exclude a (Spring MVC)Controller, or is this the wrong way to approach this?
By the way, I also tried to annotate tho at the method:
#PreAuthorize("hasRole('IS_AUTHENTICATED_ANONYMOUSLY')")
Why don't you try permitAll instead?
<intercept-url pattern="/index.jsp" access="permitAll" />

How do I Resolve an application URL from a background thread in ASP.NET MVC?

The app splits off into two threads; the main web app and a secondary thread used for asynchronous event handling. The secondary thread receives an event where it needs to send an email with a fully qualified URL of the main app (plus additional route arguments) inside it.
Eg. http://Server.com/App/RouteData?AdditionalArguments
Of course the background thread does not have the luxury of using HttpContext.Current to resolve a Url, because there's no request. No HttpRequest, no HttpContext...
I discovered that most of the methods ASP.NET (even with MVC) uses to build URLs rely on HttpContext. Does there exist a way to build a fully qualified application URL in ASP.NET without using HttpContext or any of it's derivatives?
I'm looking for a thread-safe method like:
UrlHelper.GetApplicationtUrl()
Any Ideas? Your suggestions are much appreciated.
I had this exact problem. I ended up storing the url in the web.config file. I did mine like so:
<appSettings>
<!-- Urls -->
<add key="AppDomain" value="http://localhost:1273/" />
<add key="ConfirmUrl" value="http://localhost:1273/Auth/Confirm/?code={0}" />
</appSettings>
and called it like this in the service layer:
string confirmUrl = string.Format(ConfigurationManager.AppSettings["ConfirmUrl"], confirmCode);
If you can't just use the configuration file, when creating the Thread, use the ThreadStart delegate to provide the base information you need to the new thread.

Resources