[keycloak-dev] Access code length

[Stian Thorgersen]

Thanks for suggestions.

Now the code looks like:

  xuAZKwa2EkX_WBHp-LPHXXQYKc0.YjUzMmRhODAtZTFkZS00MTZjLTk1ZjktNGJlMGYxMDNmNGE2LTE0MDY2Mjc0NDg4NjE

This consists of two parts (both Base64 url-encoded):

  <Hashed Signature>.<ID>

The hashed signature is the SHA-1 digest of the SHA256withRSA signature of ID + timestamp + action

To verify a code the details for the AccessCode is retrieved using the first part of the code. Then the hashed signature is created from the stored details and compared to the first part of the code.

This provides both brute force protection, and also protects against UUID collisions (which should be very unlikely, but still).

----- Original Message -----
> From: "Bill Burke" <bburke at redhat.com>
> To: keycloak-dev at lists.jboss.org
> Sent: Monday, 28 July, 2014 7:48:34 PM
> Subject: Re: [keycloak-dev] Access code length
> 
> Smaller hash (sha1 or md5) is really the only thing you can do beyond
> removing the JWS header.  But, if you're going to do that, might as well
> just ditch the signature altogether.  If anything, over time, we'd be
> using larger hashes not smaller ones.
> 
> HMAC is smaller, but that would require a realm secret key.
> 
> With a HMAC header and computing the JWS of:
> 
>          String uuid = UUID.randomUUID().toString();
> 
> 
> 156 chars for HMAC 512 secret key
> 113 chars for HMAC 256 key
> 99 chars for HMAC SHA1 (We'd have to implement SHA1 which would be easy
> to add)
> 
> 
> Header is 20+ characters.  uuid is 36 chars.
> 
> My bet is that Google is using HMAC + SHA-1
> 
> 
> 
> 
> 
> On 7/28/2014 12:59 PM, Marek Posolda wrote:
> > Something, which comes to my mind:
> >
> > * Omit the header part with algorithm info and instead assume that code
> > is always signed with rsa-256 (or whatever different algorithm)
> >
> > * Is it really needed to have System.currentTimeMillis() info here? I
> > wonder if this timestamp info can be extracted to server side as well,
> > so when code is later looked-up, you will also have info about the time
> > when it was created without the need of pass this info as part of the code
> >
> > * It seems that the biggest part (around 170 characters) is signature. I
> > wonder that it can be reduced by either:
> > ** Use not so strong algorithm to sign the code (not sure if it's
> > security issue as code is one time usage though and it's better than
> > avoid signature at all...)
> > ** Actually it's keycloak itself, which verifies signature. So I wonder
> > if instead of passing the whole signature around, it can just pass the
> > hash of signature (just few characters). Then later during verification
> > phase, it has access to whole accessCode content, so it can sign the
> > content again and compute again the hash of signature and compare it
> > with the hash passed from code parameter.
> >
> > Marek
> >
> > On 28.7.2014 18:14, Stian Thorgersen wrote:
> >> I've made the changes to access code, but I'm still not happy with the
> >> length.
> >>
> >> The old access code containing all the details (including access token and
> >> id token) was at least 1000 characters long (with more roles and id token
> >> details that would have been even bigger!).
> >>
> >> The new access code length is 259 characters. I think this is still to big
> >> and would like to reduce it. The way I'm currently creating it is taking
> >> the ID (UUID.randomUUID() + System.currentTimeMillis()) and using
> >> JWSBuilder to add a signature. Example:
> >>
> >>     eyJhbGciOiJSUzI1NiJ9.MTVhMjFiNWMtNjcxYy00YzYyLWEyOGMtNmFmMzE4YzE0NjkzMTQwNjU2MzMwNzQyMw.djN2v1Egz1Fime-2bVvYyv529KNMChYh3e0rY5UU4TImv53ppwNhrmWaZRNzO0xLULRC8IeixrH6K3nR1cMyHZb9Ef8Vr46YvNmE8Q_qac821GAWMNzEKjYcXt87AhSSz8BY54CcbKIPLfxFC8smI-KDA1Mr5SrJ_Ch-zdLDbdM
> >>
> >> I checked Google it's just:
> >>
> >>     4/YJcHpG7-DcoVcwNH_8D-NDQKcP_Y.MijWXLPgpOwToiIBeO6P2m9c72_7jgI
> >>
> >> Suggestions welcome ;)
> >> _______________________________________________
> >> keycloak-dev mailing list
> >> keycloak-dev at lists.jboss.org
> >> https://lists.jboss.org/mailman/listinfo/keycloak-dev
> >
> > _______________________________________________
> > keycloak-dev mailing list
> > keycloak-dev at lists.jboss.org
> > https://lists.jboss.org/mailman/listinfo/keycloak-dev
> >
> 
> --
> Bill Burke
> JBoss, a division of Red Hat
> http://bill.burkecentral.com
> _______________________________________________
> keycloak-dev mailing list
> keycloak-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/keycloak-dev
>