CSRFGuard 2.2 ChangeLog

Jump to: navigation, search



OWASP CSRFGuard 2.2 offers several advantages over previous releases. The purpose of this document is to briefly describe these changes and why they were required for a more robust and flexible implementation of a portable unique request token mechanism.

Detailed Change List

OWASP CSRFGuard Comes With a Custom JSP Tag Library For Token Insertion

One of the coolest updates to the OWASP CSRFGuard is the inclusion of a JSP tag library that developers can use to place the CSRF token in their HTML. This is especially useful for Ajax-y, Web 2.0 application that contain boat loads of background requests. In previous releases, the ability to actually use CSRFGuard was dependent on whether or not it could successfully put the token in the HTML response. With the introduction of the custom JSP tag library, this is no longer the case. Developers can use the JSP tag library to place the token in the appropriate locations of all HTML and JavaScript content. Such a feature should aide in the use of OWASP CSRFGuard in more modern and edge-y HTML and JavaScript based J2EE web applications.

While I would love to take credit for this idea, although in good conscience I cannot. A buddy of mine at the OWASP San Jose conference suggested the idea after I gave my talk about CSRFTester and CSRFGuard. I must say - it is an honor to be around so many smart people in this community!

JavaScript Handler Processes All DOM Elements

Previous releases of CSRFGuard iterated over a set of DOM elements known to contain an 'href' or 'src' attribute that required the token. This approach is extremely flawed for two reasons: 1) there was at least one tag identified that was not known and 2) the release of the HTML 5 spec further complicates the management of known elements. Rather than implementing a black-list of dangerous tags, the JavaScript handler now iterates over and processes all tags in the DOM. Any tag containing a 'href' or 'src' attribute will have the token appended.

Developer Can Define Entry Point Pages

Consider the case when we have an unprotected page that links to a resource protected by CSRFGuard. That link on the unprotected page will not have the required unique request token to pass the CSRF defense. As a result, there were a number of false CSRF attacks detected in previous releases. To address this issue, the latest version of CSRFGuard allows the developer to define "entry point" pages. These are pages for which the existence of a token should never be validated but for which the token must always be inserted.

Developer Can Define Unprotected Pages

There are a number of scenarios why certain pages or resources should not be protected by OWASP CSRFGuard. In fact, fine-grain specification of protected resources can help overcome the overhead introduced through the use of CSRFGuard. Developers using the latest version of CSRFGuard can now specify pages that should not be protected. These pages are not validated for the existence of the token and they do not get the token inserted in their HTML.

Optionally Verify Token If Parameters Exist in the Request

Previous versions of CSRFGuard would validate every single request sent to a protected resource. Consider the case when you click the "Refresh" button or when you click a book mark. When these actions occur, no valid token will exist in the GET request. As a result, CSRFGuard will incorrectly detect an attack. To counter this issue, developers can configure OWASP CSRFGuard to only validate the request if there are one or more parameters in the request. This is achieved through the use of the "org.owasp.csrfguard.ParameterlessValidation" directive. If set to false, CSRFGuard will only validate if there is one or more parameters. If set to true, CSRFGuard will mark a parameterless request for a protected resource as a possible attack. Please understand the potential consequences of this configuration. There are many web applications (Ex. Drupal, http://www.drupal.org) that invoke server-side actions using parameterless requests. If your application has this behavior, then I would highly recommend you consider validating all requests by setting "ParameterlessValidation" to true. If your web application(s) does not invoke server-side actions using parameterless requests, consider setting "ParameterlessValidation" to false to increase overall user experience.

Developer Can Specify Initial Landing Page Upon Token Creation

Previous releases of OWASP CSRFGuard had a very innocuous bug that allowed for a one-time CSRF attack against a particular session. In previous versions, when the first request came in for a session, the token was created and the request (valid or forged!) would be allowed through. As a result, there existed the possibility of performing a CSRF attack against an unauthenticated and session-less user. To counter this particular deficiency, developers can specify a "landing page" for the initial request when the CSRF token is being created. Rather than allowing the initial request through after token creation, the user can be optionally redirected to a safe landing page.

Fixed a Number of Code Quality Bugs

The OWASP CSRFGuard 2.0 release was a bit rushed as I was trying to get all of the latest features implemented in time for the OWASP San Jose conference last year. As a result, a number of code quality bugs slipped by me. One such issue was an endless loop in the FilterOutputStream object. To address this issue, three major changes were implemented for this release - the appropriate use of a unit testing framework (jUnit), the use of code quality static analyzer(s) (FindBugs, PMD), and a more stringent peer review. OWASP CSRFGuard 2.0 users are strongly encouraged to update to the latest release.

The RegExHandler is Deprecated

Sorry guys - that class was a complete hack since day one of CSRFGuard. That "buildLinkParameters" method is quite possibly the ugliest code I've written since working on a particular VBA application. There are so many better and more useful options that I've decided not to spend my limited energy improving this response handler. If someone would like to lead the development of the RegExHandler, shoot me an email.