Re: [OAUTH-WG] Clarification of "client application consisting of multiple components"

[Kris Selden <[e-mail]>]

Thanks,

I understand that the spec leaves it ambiguous, I read the long thread on multiple components single client ID.

I wanted to point out that this will be common despite this. Hopefully maybe to encourage FB to start an extension clarify this.

Thanks for your help, I had overlooked using graph.facebook.com/app?access_token=your-token as a way to verify the client id.

On Apr 5, 2012, at 6:19 PM, "matake, nov" <[e-mail]> wrote:

> OAuth Core spec doesn't define those cases, so OAuth WG ML isn't the place to report this issue though.
> * how to handle an app which consists both client-side and server-side components.
> * how to use OAuth for login
> 
> I already reported this issue to
> * apple
> * facebook
> * foursquare
> * pinterest
> * yapp
> 
> Not all of them are understanding this issue though..
> 
> 2012/4/6 matake, nov <[e-mail]>
> Let me describe the details first.
> 
> FB iOS SDK delegates the authorization step to the official FB iOS app via "fbauth://authorize" custom schema URL.
> (If the official app isn't available on the device, it just open m.facebook.com authorization page using Safari)
> 
> After the end-user approved the client access, FB server returns token and code to the official FB app.
> Then FB app passes token and code to the original app via the app's custom schema (fb123456://authorize, numeric part is FB app_id of the app).
> So if the app has its own server-side component, it should pass the code to the server, not the access token.
> (ref. http://www.thread-safe.com/2012/01/problem-with-oauth-for-authentication.html)
> 
> However, current FB iOS SDK just ignores the code and almost all developers seems not using code.
> Most apps are sending access tokens to their own server-side component. (foursquare, pinterest, yapp etc.)
> So the vulnerability John Bradley described before exists in many iOS app using FB iOS SDK now. (maybe Android apps too)
> http://www.thread-safe.com/2012/02/more-on-oauth-implicit-flow-application.html
> 
> There are 2 solutions for this issue.
> 
> First is modify FB iOS SDK and make code accessible from your app.
> I'm already talking with FB guys about it, so I hope they change their code by themselves.
> For now, you can use my fork of FB SDK too.
> https://github.com/nov/facebook-ios-sdk
> 
> If 1st one is hard for you (because you cannot remove legacy version of your app from the users device etc), you can also use FB Graph API endpoint to verify access token audience.
> Send a GET request to graph.facebook.com/app?access_token=your-token, then you'll get application info to which the token is established.
> 
> 2012/4/6 Kristofor Selden <[e-mail]>
> How do I deal with this?
> https://twitter.com/#!/nov/status/187895781011890176
> 
> My assumption is after getting the user to authorize the client via the FB SDK on the iPhone app, one would send the authorization code (not the access token) back to the server via HTTPS where it would just get a new token using the client id+secret and the authorization code, given that the server client id is the same as iPhone apps client id.
> 
> Is this correct?
> 
> With mobile apps, having multiple components use the same client id is going to be common regardless if that is less than ideal.  It is common to have a native mobile client component paired with a server component both using FB social graph using the same client_id.
> 
> For many startups FB won the social graph war and we just want to leverage that and focus on our offerings.
> 
> Thanks,
> Kris
> 
> On Mar 11, 2012, at 10:43 AM, nov matake wrote:
> 
>> I understood.
>> Thanks.
>> 
>> --
>> nov
>> 
>> On Mar 12, 2012, at 2:30 AM, Eran Hammer <[e-mail]> wrote:
>> 
>>> That use case was removed from the specification. Either way, it is up to the authorization server to decide which registration options to offer the client if they make such a grant type available in the future, and how it will apply the security policies. IOW, those proposing such an extension in the future will have to figure out how this should be handled.
>>> 
>>> EH
>>> 
>>>> -----Original Message-----
>>>> From: nov matake [mailto:[e-mail]]
>>>> Sent: Sunday, March 11, 2012 10:21 AM
>>>> To: Eran Hammer
>>>> Cc: nov matake; [e-mail] WG
>>>> Subject: Re: [OAUTH-WG] Clarification of "client application consisting of
>>>> multiple components"
>>>> 
>>>> So what is the usecase of response_type=token%20code ?
>>>> I thought, in that usecase, token was for the client's client-side component,
>>>> code was for the client's server-side component, and both of them have the
>>>> same client_id.
>>>> 
>>>> --
>>>> nov
>>>> 
>>>> On Mar 12, 2012, at 12:57 AM, Eran Hammer <[e-mail]> wrote:
>>>> 
>>>>> If you have two components each with different security profile, you must
>>>> assign each a different client_id. Otherwise, there is no way to enforce the
>>>> rest of the spec's security requirements.
>>>>> 
>>>>> EH
>>>>> 
>>>>>> -----Original Message-----
>>>>>> From: [e-mail] [mailto:[e-mail]] On
>>>>>> Behalf Of nov matake
>>>>>> Sent: Sunday, March 11, 2012 8:25 AM
>>>>>> To: [e-mail] WG
>>>>>> Subject: [OAUTH-WG] Clarification of "client application consisting
>>>>>> of multiple components"
>>>>>> 
>>>>>> Hi,
>>>>>> 
>>>>>> I just found this sentence in the latest draft.
>>>>>> 
>>>>>> Does it mean "an application consisting of server-side and
>>>>>> client-side component (eg. foursquare iPhone app) MUST have separate
>>>>>> client_id for each component" ?
>>>>>> Or can I image something like Facebook is doing right now? (register
>>>>>> each component for a single client_id separately)
>>>>>> 
>>>>>> ==
>>>>>> A client application consisting of multiple components, each with its
>>>>>> own client type (e.g. a distributed client with both a confidential
>>>>>> server-based component and a public browser-based component), MUST
>>>>>> register each component separately as a different client to ensure
>>>>>> proper handling by the authorization server.  The authorization
>>>>>> server MAY provider tools to manage such complex clients through a
>>>> single administration interface.
>>>>>> ==
>>>>>> 
>>>>>> --
>>>>>> nov <[e-mail]>
>>>>>> _______________________________________________
>>>>>> OAuth mailing list
>>>>>> [e-mail]
>>>>>> https://www.ietf.org/mailman/listinfo/oauth
>>>>> _______________________________________________
>>>>> OAuth mailing list
>>>>> [e-mail]
>>>>> https://www.ietf.org/mailman/listinfo/oauth
>>> _______________________________________________
>>> OAuth mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/oauth
>> _______________________________________________
>> OAuth mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/oauth
>> 
> 
> 
>