Re: W3C's Enrollment Specification

On 6 Aug 2011, at 12:06, Anders Rundgren wrote:

> On 2011-08-06 11:38, Henry Story wrote:
>> On 6 Aug 2011, at 10:04, Anders Rundgren wrote:
>>> http://lists.w3.org/Archives/Public/public-html/2011Aug/0033.html
>>> 
>>> W3C is like PKIX working with the idea of upgrading existing schemes
>>> rather than starting with a requirement specification and see where
>>> that leads you.
>>> 
>>> I don't think W3C's revised <keygen> will go anywhere because a 2-phase
>>> protocol doesn't really cut it.  Apple's already deployed scheme for iPhone
>>> is considerably more powerful and user-friendly.
>> 
>> The MD5 situation can be mitigated by the server using a time based challenge.
>> This can reduce the attack surface to a few minutes. I doubt md5 is that bad.
>> But better security would be better of course.
> 
> Yes, I think this particular "problem" is irrelevant and doesn't need solving.
> 
>> I wrote this up the different ways of creating certificates here
>> 
>> http://www.w3.org/wiki/Foaf%2Bssl/Clients#Support_for_easy_creation_of_certificates
>> 
>> What I am still not clear about is what could go wrong. I thought I had understood 
>> that for a while, but I realised I am not clear about that. After all a public 
>> certificate is no use if you do not have the private key corresponding to the public key
>> published in the certificate. So even if someone took the public key generated by the browser
>> there is not much they could do with it.
>> 
>> Can you fill be in again here? I feel like there is something I am missing here, and I would
>> like to fill in the whole in the wiki above.
> 
> That's absolutely correct, you get nowhere with a certificate without
> the matching private key.  In fact, the PoP (Proof-of-Possession) schemes
> featured in enrollment schemes are redundant. Attestations of the kind
> there are in ETSI/3GPP and TCG protocols OTOH, actually fills a purpose
> since they identify the key-container.

Ah you mean the crypto key, or the telephone keychain?

I suppose that is nice if you want to tie someone to a device, but problematic if you want more self asserted identity for privacy reasons. 

It could be that there is space for both schemes.
> 
> 
>> By the way I don't see how what Apple is doing could have a better user interface.
>> The user interface for keygen is: click a button. Unless they move to mind reading...
> 
> Well, user-interface is just one aspect but if we concentrate on that one, Apple's
> solution eliminates the confusing strong/weak button.

Agree, the strong/weak button should just be a hint. Mind you it could be thought of that
way. Opera gives way too many options there, Other browsers give much fewer. So it would
not be impossible to have a browser just give none. Work to eliminate that button to
everyone's satisfaction would certainly be welcome.

> 
> Anders
> 
>> 
>> Henry
>> 
>> 
>>> 
>>> Anders
>>> 
>> 
>> Social Web Architect
>> http://bblfish.net/
>> 
>> 
>> 
> 

Social Web Architect
http://bblfish.net/

Received on Saturday, 6 August 2011 10:23:30 UTC