Does vulnerability exist when using XHR with GET method and custom anti-CSRF HTTP header?

Question

Imagine button after clicking which browser sends XHR http request with GET method. Characteristics:

• after executing request sensitive action is performed
• sensitive information is sent in GET parameters
• HTTP headers X-Requested-With: XMLHttpRequest and X-CSRF-Token with large enough PRG token are sent and are necessary to have the action executed
• option to Copy link location isn't accessible when rightclick the button as it's a form

Normally XHR is used with GET but here it's used with GET. It's the only difference from normal XHR requests.

Does this design lead to any security vulnerability?

Edit: Some answerers and commenters tell that:

• User may copy or bookmark url. However, the only place from where user may copy url is the source code of page because this url isn't exposed in browser address bar, rightclick menu because of button is a form). Also parameters are added into request by browser itself so user can't copy full url from source code.
• User may copy part of browser window. However, url isn't exposed in browser window
• It may become part of Referer header. However, it's XHR and doesn't change url in Referer header of next requests

So the only valid vulnerability seems to be that GET parameters may be logged in webserver logs (they won't be logged at proxy because of using TLS). If those logs are accessible to attacker, then he will have sensitive information.

Answer 1

I wouldn't rely on this design. The HTTP specification states:

Implementors should be aware that the software represents the user in their interactions over the Internet, and should be careful to allow the user to be aware of any actions they might take which may have an unexpected significance to themselves or others.

In particular, the convention has been established that the GET and HEAD methods SHOULD NOT have the significance of taking an action other than retrieval. These methods ought to be considered "safe". This allows user agents to represent other methods, such as POST, PUT and DELETE, in a special way, so that the user is made aware of the fact that a possibly unsafe action is being requested.

Naturally, it is not possible to ensure that the server does not generate side-effects as a result of performing a GET request; in fact, some dynamic resources consider that a feature. The important distinction here is that the user did not request the side-effects, so therefore cannot be held accountable for them.

More generally speaking I would never recommend using CSRF tokens in GET requests since they may become available in various log files (vulnerability if the anticsrf token is not properly disposed of after one use.), they may become part of the referrer header, users may bookmark them and this will lead to either a security leak or to invalid bookmarks. This may have negative impact on the SEO too. Etc, etc.

So I think that GET+CSRF+ANY_OTHER_SENSITIVE_DATA=bad idea.

Answer 2

• Issue 1: Sensitive data in the URL.
• Issue 2: CSRF token in URL Anything you put in the URL will be logged in browser history, proxy logs, webserver logs ....
• Issue 3: You can't prevent user from viewing the link/url by disabling the right click options, anything you sent to browser will be accessible to the user, one way or other

DO NOT USE GET for posting data and requests which change the state.

I don't see any problem in putting the CSRF token in the header, additional things you need to consider are not opening your server to HTML5 CORS requests . If you need to allow CORS then there additional things you would need to consider.