Open Stack

You're quite probably correct - going through the OWASP threat list in more
detail is on my TODO. That was just off the top of my head as something
that has me concerned but I've not investigated it thoroughly.

On 17 September 2014 14:15, Adam Young <ayoung at redhat.com> wrote:

>  On 09/16/2014 08:56 PM, Richard Jones wrote:
>
> CORS for all of OpenStack is possible once the oslo middleware lands*, but
> as you note it's only one of many elements to be considered when exposing
> the APIs to browsers. There is no current support for CSRF protection in
> the OpenStack APIs, for example. I believe that sort of functionality
> belongs in an intermediary between the APIs and the browser.
>
>
> Typically, CSRF is done by writing a customer header.  Why wouldn't the
> -X-Auth-Token header qualify?  Its not a cookie, automatically added.  So,
> CORS support would be necesary for horizon to send the token on a request
> to Nova, but no other site would be able to do that.  No?
>
>
>
>
>      Richard
>
>  * https://review.openstack.org/#/c/120964/
>
> On 17 September 2014 08:59, Gabriel Hurley <Gabriel.Hurley at nebula.com>
> wrote:
>
>> This is generally the right plan. The hard parts are in getting people to
>> deploy it correctly and securely, and handling fallback cases for lack of
>> browser support, etc.
>>
>> What we really don't want to do is to encourage people to set
>> "Access-Control-Allow-Origin: *" type headers or other such nonsense simply
>> because it's too much work to do things correctly. This becomes especially
>> challenging for federated clouds.
>>
>> I would encourage looking at the problem of adding all the necessary
>> headers for CORS as an OpenStack-wide issue. Once you figure it out for
>> Keystone, the next logical step is to want to make calls from the browser
>> directly to all the other service endpoints, and each service is going to
>> have to respond with the correct CORS headers
>> ("Access-Control-Allow-Methods" and "Access-Control-Allow-Headers" are
>> particularly fun ones for projects like Glance or Swift). A common
>> middleware and means of configuring it will go a long way to easing user
>> pain and spurring adoption of the new mechanisms. It will help the Horizon
>> team substantially in the long run to do it consistently and predictably
>> across the stack.
>>
>> As a side-note, once we're in the realm of handling all this sensitive
>> data with the browser as a middleman, encouraging people to configure
>> things like CSP is probably also a good idea to make sure we're not loading
>> malicious scripts or other resources.
>>
>> Securing a browser-centric world is a tricky realm... let's make sure we
>> get it right. :-)
>>
>>      - Gabriel
>>
>> > -----Original Message-----
>> > From: Adam Young [mailto:ayoung at redhat.com]
>> > Sent: Tuesday, September 16, 2014 3:40 PM
>> > To: OpenStack Development Mailing List
>> > Subject: [openstack-dev] [Keystone][Horizon] CORS and Federation
>> >
>> > Phase one for dealing with Federation can be done with CORS support
>> solely
>> > for Keystone/Horizon  integration:
>> >
>> > 1.  Horizon Login page creates Javascript to do AJAX call to Keystone 2.
>> > Keystone generates a token 3.  Javascript reads token out of response
>> and
>> > sends it to Horizon.
>> >
>> > This should support Kerberos, X509, and Password auth;  the Keystone
>> team
>> > is discussing how to advertise mechanisms, lets leave the onus on us to
>> solve
>> > that one and get back in a timely manner.
>> >
>> > For Federation, the handshake is a little more complex, and there might
>> be a
>> > need for some sort of popup window for the user to log in to their home
>> > SAML provider.  Its several more AJAX calls, but the end effect should
>> be the
>> > same:  get a standard Keystone token and hand it to Horizon.
>> >
>> > This would mean that Horizon would have to validate tokens the same way
>> > as any other endpoint.  That should not be too hard, but there is a
>> little bit of
>> > "create a user, get a token, make a call" logic that currently lives
>> only in
>> > keystonemiddleware/auth_token;  Its a solvable problem.
>> >
>> > This approach will support the straight Javascript approach that
>> Richard Jones
>> > discussed;  Keystone behind a proxy will work this way without CORS
>> > support.  If CORS  can be sorted out for the other services, we can do
>> straight
>> > Javascript without the Proxy.  I see it as phased approach with this
>> being the
>> > first phase.
>> >
>> >
>> >
>> >
>> >
>> > _______________________________________________
>> > OpenStack-dev mailing list
>> > OpenStack-dev at lists.openstack.org
>> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>> _______________________________________________
>> OpenStack-dev mailing list
>> OpenStack-dev at lists.openstack.org
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>
>
>
> _______________________________________________
> OpenStack-dev mailing listOpenStack-dev at lists.openstack.orghttp://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140917/6baf1d49/attachment.html>