Skip to content

Commit

Permalink
Don't check authenticity tokens for any AJAX requests
Browse files Browse the repository at this point in the history
  • Loading branch information
Ross Kaffenburger and Bryan Helmkamp authored and wycats committed Apr 15, 2009
1 parent 3c11876 commit 256b0ee
Show file tree
Hide file tree
Showing 3 changed files with 10 additions and 6 deletions.
2 changes: 2 additions & 0 deletions actionpack/CHANGELOG
Expand Up @@ -7,6 +7,8 @@

* Fixed that redirection would just log the options, not the final url (which lead to "Redirected to #<Post:0x23150b8>") [DHH]

* Don't check authenticity tokens for any AJAX requests [Ross Kaffenberger/Bryan Helmkamp]

* Added ability to pass in :public => true to fresh_when, stale?, and expires_in to make the request proxy cachable #2095 [Gregg Pollack]

* Fixed that passing a custom form builder would be forwarded to nested fields_for calls #2023 [Eloy Duran/Nate Wiger]
Expand Down
Expand Up @@ -81,12 +81,13 @@ def verify_authenticity_token

# Returns true or false if a request is verified. Checks:
#
# * is the format restricted? By default, only HTML and AJAX requests are checked.
# * is the format restricted? By default, only HTML requests are checked.
# * is it a GET request? Gets should be safe and idempotent
# * Does the form_authenticity_token match the given token value from the params?
def verified_request?
!protect_against_forgery? ||
request.method == :get ||
request.xhr? ||
!verifiable_request_format? ||
form_authenticity_token == params[request_forgery_protection_token]
end
Expand Down
11 changes: 6 additions & 5 deletions actionpack/test/controller/request_forgery_protection_test.rb
Expand Up @@ -151,14 +151,10 @@ def test_should_not_allow_api_formatted_delete_sent_as_multipart_form_without_to
delete :index, :format => 'xml'
end
end

def test_should_allow_xhr_post_without_token
assert_nothing_raised { xhr :post, :index }
end
def test_should_not_allow_xhr_post_with_html_without_token
@request.env['CONTENT_TYPE'] = Mime::URL_ENCODED_FORM.to_s
assert_raise(ActionController::InvalidAuthenticityToken) { xhr :post, :index }
end

def test_should_allow_xhr_put_without_token
assert_nothing_raised { xhr :put, :index }
Expand All @@ -168,6 +164,11 @@ def test_should_allow_xhr_delete_without_token
assert_nothing_raised { xhr :delete, :index }
end

def test_should_allow_xhr_post_with_encoded_form_content_type_without_token
@request.env['CONTENT_TYPE'] = Mime::URL_ENCODED_FORM.to_s
assert_nothing_raised { xhr :post, :index }
end

def test_should_allow_post_with_token
post :index, :authenticity_token => @token
assert_response :success
Expand Down

36 comments on commit 256b0ee

@Manfred
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yehuda, can you explain this commit? Is this associated with a Lighthouse ticket somehow?

@corasaurus-hex
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why was this removed?

@jasonroelofs
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Um, wasn’t half the point of Auth Tokens security against rogue AJAX requests? What is the point of this commit?

@norbert
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, this seems odd.

@wycats
Copy link
Member

@wycats wycats commented on 256b0ee Apr 16, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@jameskilton rogue Ajax requests cannot produce CSRF attacks, so this was just an annoyance to people hand-writing JS.

@carllerche
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

jameskilton: If you can make a rogue AJAX request, you have bigger problems than CSRF attacks ;)

@labria
Copy link
Contributor

@labria labria commented on 256b0ee Apr 17, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

At last, no dumb tokens to generate by hand =)

@wycats
Copy link
Member

@wycats wycats commented on 256b0ee Apr 17, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@labria at last… a positive comment on this change! W00H00!

@lucashungaro
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Very nice indeed. :)

@NZKoz
Copy link
Member

@NZKoz NZKoz commented on 256b0ee Apr 17, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The same original policy protects ajax requests. you can’t make browsers submit forms or make requests with custom headers.

This should be safe until browsers drop the same origin policy, which is (one hopes) forever.

@adrianpacala
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why thank you kind sir. No more ugly inline JS only for that one pesky line of auth token.

@gbuesing
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is also a big win for cacheability, because we can remove user-specific authenticity tokens from ajaxified links and forms that don’t need to support a non-ajax fallthrough.

Very nice!

@bit-forge
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Excellent! This was a pain when implementing unobtrusive JS…

@tilsammans
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don’t understand this commit. What’s stopping an attacker who is compiling a browser herself and making different-origin a reality? Now the security of Rails apps is tied to ‘hopes’ of third parties doing the right thing?

@samleb
Copy link
Contributor

@samleb samleb commented on 256b0ee Apr 17, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@tilsammans: I believe CSRF protection is only useful to prevent users from doing cross-site requests without knowing it (i.e. submitting a form whose

action
is on another site where he’s authenticated).
If one compiles its own browser making different-origin a reality as you suggested, he will be the only one in danger. Am I missing something here ?

@samleb
Copy link
Contributor

@samleb samleb commented on 256b0ee Apr 17, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry about the preformatted tag above… having previews or comment edition would be such a nice feature ;)

@RStankov
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Really great addition, will save me a lot unnecessary hacking :)

@tilsammans
As far as I know protected_from_forgery, is for securing the frontend users, from other users/sites/scripts, who tried to pretend to be the frontend user. I mean that a person who is compiling a browser himself, for example, could only take his session data, witch is his. This is the same as that I could change the html with FireBug and fire some forms, run ajax queries, and other things from every site.

@milodurden
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So what happens when someone posts a CSRF request with the header ‘X-Requested-With’ set to ‘XMLHttpRequest’ ?

@NZKoz
Copy link
Member

@NZKoz NZKoz commented on 256b0ee Apr 18, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@mildodurden: You can’t do that with browsers as they exist currently. If you can prove otherwise, please let us know and we can revert this.

@NZKoz
Copy link
Member

@NZKoz NZKoz commented on 256b0ee Apr 18, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@tilsammans: If the attacker is executing custom browsers on the user’s computer, there’s nothing we can do about it. They could also be running keyloggers, screenshotters, botnets etc.

This is for protecting against CSRF attacks, not a silver bullet for all possible attacks ever inventable.

@wycats
Copy link
Member

@wycats wycats commented on 256b0ee Apr 18, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For further details on CSRF, check out this great Stanford paper:

XMLHttpRequest’s popularity has increased recently with more sites implementing AJAX interfaces. Sites can defend against CSRF by setting a custom header via XMLHttpRequest and validating that the header is present before processing state-modifying requests. Although effective, this defense requires sites to make all state-modifying requests via XMLHttpRequest, a requirement that prevents many natural site designs.

@arthurgeek
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Any chance to see this change on 2.3 branch? :)

@arthurschreiber
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1 for a backport! :)

@josevalim
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1 for a backport!

@samgranieri
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1 for a backport

@corasaurus-hex
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You could backport it in a backwards compatible way. Check it if it’s included, otherwise ignore it.

@ultrasaurus
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1 for for 2.3

@ultrasaurus
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just saw binary42’s response. There reason to backport it, is that new code is still being written for Rails and this is a stumbling block when you want to make requests from Ajax, desktop client, or Flash. I don’t see this as a policy change, so much as a bug fix. (But I’m new to Rails, maybe I’m missing something)

@tessro
Copy link
Contributor

@tessro tessro commented on 256b0ee Apr 20, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What about local XSS attacks? With unescaped output, something akin to the MySpace worm is enabled by this change.

Granted, Rails coders always remember to escape output! :) But was the intent to remove a layer of defensive programming?

@ultrasaurus
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

-1 for 2.3 (revised vote)

@binary42 after thinking about this more this morning, I realized that changing the behavior in a dot release is probably unwise. If there is an existing app depending on this behavior with an open cross-domain.xml and using cookie-based authentication, then a malicious flash app hosted on another site could make requests on behalf of a user without them being aware of it.

In terms of the “5 minute” fix… I’m still struggling to find the right few lines of code to write. I thought it work would to add the following to my controller:

def verify_authenticity_token
  request.xhr?      ||
  verified_request? || 
  raise(ActionController::InvalidAuthenticityToken)
end

but I still get the exception. I’m sure I’m missing something basic. Is the code change I need to make documented somewhere?

Thanks!

@gbuesing
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ultrasaurus for a quick 2.3 patch, check out this gist: http://gist.github.com/98660 . You can put this an initializer and it should work

@arthurschreiber
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@paulrosania: If your users are able to execute JavaScript through a XSS, this authenticity token won't be able to prevent CSRFs at all. It would be a snap to fetch the authenticity token from a page which has a form on it...

@everyone else who wants to see this in 2.3: http://gist.github.com/99635 <- a simple initializer like the one of gbuesing, but not using class eval.

@NZKoz
Copy link
Member

@NZKoz NZKoz commented on 256b0ee Apr 22, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@paulrosania: In fact the twitter worm did exactly what NoKarma is describing. XSS is the stack-smashing of web attacks, if you're attacked you're done. Nothing can save you

@vrybas
Copy link
Contributor

@vrybas vrybas commented on 256b0ee May 4, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ultrasaurus:
skip_before_filter :verify_authenticity_token, :only => [:aircrafts_by_manufacturer]
?

@NZKoz
Copy link
Member

@NZKoz NZKoz commented on 256b0ee May 4, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@vrybas: instead of explicitly skipping the filter you should do:

protect_from_forgery :except=>[:aircrafts_by_manufacturer]

I'm also reasonably sure that you could do something along the lines of

protect_from_forgery :if=>:some_method_which_returns_true_when_ajax_or_whatever

@jdwyah
Copy link

@jdwyah jdwyah commented on 256b0ee Jul 17, 2009

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Isn't it pretty easy to add authenticity_tokens to all ajax requests? ie these 7 lines of code? http://gist.github.com/149110

Please sign in to comment.