Re: [jose] DISCUSS: Nonce/Timestamp parameter

[Mike Jones <[e-mail]>]

As editor, I'm going to make the observation this is the one poll question where the results are not clear enough for it to be obvious what I should do.  There were many people who made comments about the question being unclear and the intended use and meaning of the potential field or fields unclear.

Unless you feel differently, Karen and Jim, I believe that the best course at this point is for me to add nothing to the specs as a result of this poll question, but for the working group to try to make decisions on much less ambiguous proposals (should any be made) than the one in the poll question.

                                                            -- Mike

From: [e-mail] [mailto:[e-mail]] On Behalf Of Jim Schaad
Sent: Tuesday, August 28, 2012 8:38 PM
To: 'Dick Hardt'; [e-mail]
Cc: [e-mail]; [e-mail]
Subject: Re: [jose] DISCUSS: Nonce/Timestamp parameter

The question from my point of view is that common fields from many structures should potentially be supported in the base specification so that they can be common rather than having each structure define them separately.  This is only an issue if one wishes them to be placed in the header structure and not in the data structure.

If one is looking at signing an unstructured data object - such as a file - then it becomes difficult to have the fields such as a time that it was signed be part of the file itself, especially if one is applying multiple signatures at different times.  This is not an issue for the token specification but could be for other uses of the signature or encryption specifications.

I would agree that "iat" is a timestamp for the purposes of this conversation.  If one wanted a formalized timestamp from a third party authority then a totally different way of going about it would be required.  I chose the term nonce or timestamp because both had been discussed in the past without any specific resolution about what is needed.

Jim


From: Dick Hardt [mailto:[e-mail]]<mailto:[mailto:[e-mail]]>
Sent: Monday, August 27, 2012 2:55 PM
To: [e-mail]<mailto:[e-mail]>
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [jose] DISCUSS: Nonce/Timestamp parameter

I was considering "iat" to be the timestamp. I was not thinking there would be an additional timestamp.

On Aug 27, 2012, at 2:13 PM, <[e-mail]<mailto:[e-mail]>> wrote:

We have exp
                https://tools.ietf.org/html/draft-ietf-oauth-json-web-token-03#section-4.1.1
and iat
                https://tools.ietf.org/html/draft-ietf-oauth-json-web-token-03#section-4.1.3
in JWT. Why do we need a timestamp?

Replay attacks of the same jwt can be mitigated through the jti claim
https://tools.ietf.org/html/draft-ietf-oauth-json-web-token-03#section-4.1.7

What do timestamp and nonce add to these?

Axel



From: Dick Hardt [mailto:[e-mail]]<mailto:[mailto:[e-mail]]>
Sent: Monday, August 27, 2012 10:23 PM
To: Brian Eaton
Cc: Nennker, Axel; Jim Schaad; [e-mail]<mailto:[e-mail]>
Subject: Re: [jose] DISCUSS: Nonce/Timestamp parameter


On Aug 27, 2012, at 1:06 PM, Brian Eaton wrote:


On Mon, Aug 27, 2012 at 12:11 PM, Dick Hardt <[e-mail]<mailto:[e-mail]>> wrote:
I have an application for JWT that is not OAuth2.

Should nonce and timestamp logic go in the application level protocol?

I prefer to NOT have the application level deal with token validity.



Having said that, nonce's are difficult to implement at scale and I have heard of many sites that don't implement them fully.

Nonce alone can't be implemented efficiently.  You have to have time stamps as well, otherwise you are stuck storing ever nonce you've ever seen, forever.

Even nonce + time stamp is challenging in distributed systems.  It adds a lot of complexity.  That complexity is sometimes merited, but not always.

Thanks for confirming my statement.

I have stopped using nonce and only use time stamps lately and have made the system relatively stateless so that a second submission of the token is ok. That may not work for everyone, but I have found that architecture to be easier to implement and scale.