Re: [OAUTH-WG] why are we signing?

["Hannes Tschofenig" <[e-mail]>]

Hi Chris, 


  _____  

From: [e-mail] [mailto:[e-mail]] On Behalf Of
Chris Messina
Sent: 09 November, 2009 17:00
To: Eran Hammer-Lahav
Cc: [e-mail]
Subject: Re: [OAUTH-WG] why are we signing?


Indeed, in the beginning of OAuth, that was one of the primary drivers that
lead us to the decision to sign everything - because of the non-SSL case... 

The OAuth signing mechanism does not sign anything, only certain parts.
Signing everything was consider difficult since the complete data stream may
not be available.  
 
There is also a TLS-PSK variant that would not require you to use
certificates if that is considered to be the problem. 
 
While it's possibly increasingly common to expect that serious developers
will go and buy an SSL cert, that may not be the case for the wider array of
hobbyist types. Now, that's not to say that they are the only audience that
needs to be addressed, but the idea was to make it harder for them to screw
up if they leaked their API calls... Clearly it turned out that the signing
bit intended to prevent against such attacks itself was too hard to
implement, and so now we're having these conversations again.


 I rather believe that the description was difficult to understand (and
hence implementers got it wrong). Once libraries are available then there is
no need for application designers to worry about these issues. 
 

Ciao
Hannes
 
At least now we have more data about what the market will bear now.


Anyway, that's my recollection. But it might also not be exactly the
explanation for what you're looking for.


Chris


On Sun, Nov 8, 2009 at 11:48 PM, Eran Hammer-Lahav <[e-mail]>
wrote:


The problem is, we are not likely to ever reach consensus on 'reasonable
security'.

For example, I don't find most cookie-based session systems reasonably
secure without SSL/TLS. Being able to sit at a coffee shop with free wifi
and a laptop and steal sessions cookies, then access people's email for the
duration the cookie is valid isn't reasonable or secure.

If you would like to try this approach, I would suggest adding next to each
option the list of common attacks still possible under those terms. It will
allow us to evaluate the added security each level of complexity brings.

EHL


> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf
> Of Brian Eaton
> Sent: Sunday, November 08, 2009 9:03 PM
> To: [e-mail]
> Subject: [OAUTH-WG] why are we signing?
>
> Hey folks -
>
> What are the use cases for cryptography in OAuth?  Why are we signing
> requests?  And how much of each request do we need to sign in order to
> be useful?
>
> As I see it, we have roughly the following menu of choices:
>
> 1) No signatures.
>     Just use bearer tokens.  Use transport layer encryption to keep
> those bearer tokens from leaking.
>
> 2) Signed tokens.
>     We could just sign a timestamp, rather than entire messages.
>
> 3) Partially signed messages.
>     We could sign just the request URL, or the request URL plus some
> parameters.
>
> 4) Fully signed messages.
>      Sign as much of the HTTP request as possible, down to the bits of
> the HTTP entity body.
>
> My guess is we need at least two out of those four choices (one with
> bearer tokens, a la OAuth 1.0 plaintext) and another with
> cryptography.  But I'm not sure whether we need to sign entire
> messages, or if we can get away with something simpler and still have
> reasonable security.
>
> Cheers,
> Brian
> _______________________________________________
> OAuth mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/oauth
_______________________________________________
OAuth mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/oauth





-- 
Chris Messina
Open Web Advocate

Personal: http://factoryjoe.com
Follow me on Twitter: http://twitter.com/chrismessina

Citizen Agency: http://citizenagency.com
Diso Project: http://diso-project.org
OpenID Foundation: http://openid.net

This email is:   [ ] shareable    [X] ask first   [ ] private