[keycloak-dev] Same Refresh token can be used multiple times to obtain access token

Stian Thorgersen sthorger at redhat.com
Wed Oct 14 00:45:16 EDT 2015


We've already discussed how to implement this. We'll add a lastRefresh for
a client session and only permit refresh/offline tokens if the iat time is
=> lastRefresh for the client session. Simple to implement. However, our
current adapters do not update the refresh tokens they use so this would
have to be updated as well. Also, it could potentially break existing
functionality. So we should probably make it a configurable option for a
realm under token settings and have it disabled by default initially. In a
future release we could then enable it by default for new installations.

On 14 October 2015 at 01:53, Bill Burke <bburke at redhat.com> wrote:

> If your application is careless with refresh tokens, there's not much
> Keycloak can do other than session management and revocation policies.
> Refresh tokens are not supposed to be distributed or shared.
>
> But, I've scheduled a jira for this.  It may get done 1.6, 1.7 at the
> latest.
>
> On 10/13/2015 4:56 PM, Kuznetsov, Mike wrote:
> > Hello Bill,
> >
> > It is my understanding that revoking the session will log out the user,
> so I'm not sure that it will address the problem. I think what we want is
> to revoke a specific refresh token after we use it once so that it cannot
> be used again.
> >
> > I'm also not sure how we would use a not-before policy, since we would
> need to update this policy every time we get a new access token / refresh
> token pair.
> >
> > Thank You,
> >
> > Mikhail Kuznetsov
> > Software Engineer
> > Hewlett Packard Enterprise
> >
> > -----Original Message-----
> > From: keycloak-dev-bounces at lists.jboss.org [mailto:
> keycloak-dev-bounces at lists.jboss.org] On Behalf Of Bill Burke
> > Sent: Tuesday, October 13, 2015 3:50 PM
> > To: keycloak-dev at lists.jboss.org
> > Subject: Re: [keycloak-dev] Same Refresh token can be used multiple
> times to obtain access token
> >
> > tokens are associated with sessions.  Log out the user's sessions, and
> the tokens are now invalid.  You can also push out a a not-before
> revocation policy.  Applications will deny any tokens issued prior to the
> not-before timestamp.  All this is doable in the admin console.
> >
> > There is an issue though with bearer-only REST services and access
> tokens.  They do not get notified of logout events as they are stateless
> and instead rely on expiration of the access token (which is why access
> tokens should be short lived).  You'd have to rely on pushing out a
> not-before revocation policy.
> >
> > IMO, because you can logout users and/or push out a revocation policy, I
> really don't see this as a high priority security issue for 1.6.  I'll see
> if I can get it in before Friday.
> >
> > On 10/13/2015 3:03 PM, Kuznetsov, Mike wrote:
> >> Hello,
> >>
> >> As advised earlier, I have put a JIRA ticket for this issue:
> >> https://issues.jboss.org/browse/KEYCLOAK-1961
> >>
> >> We are excited to start using Keycloak to support our product, but
> >> right now we are blocked due to this issue because it has been flagged
> >> by our security team when they were reviewing our design.
> >>
> >> Over the past few days, we have tried to overcome this issue by trying
> >> to revoke either the specific access token / refresh token if it is
> >> used more than once, or trying to revoke all the tokens for the user.
> >> However, we were unable to find a mechanism that would let us do this.
> >> *Do you know if there is any mechanism to revoke a specific token, or
> >> to revoke all tokens for a user? Are there any plans to implement this
> >> in the future?*
> >>
> >> **
> >>
> >> Thank You,
> >>
> >>
> >> *Mikhail Kuznetsov*
> >>
> >> Software Engineer
> >>
> >> Hewlett Packard Enterprise
> >>
> >> *From:*keycloak-dev-bounces at lists.jboss.org
> >> [mailto:keycloak-dev-bounces at lists.jboss.org] *On Behalf Of *Kamal
> >> Jagadevan
> >> *Sent:* Monday, October 12, 2015 8:26 PM
> >> *To:* stian at redhat.com; Marek Posolda
> >> *Cc:* keycloak-dev at lists.jboss.org; Jagadevan, Kamal
> >> *Subject:* Re: [keycloak-dev] Same Refresh token can be used multiple
> >> times to obtain access token
> >>
> >> Hi Stian/Marek,
> >>
> >>     Thanks for your attention in the matter.
> >>
> >> Probably you are referring to one other issue in client level, but
> >> Mike & I are referring at User level within or across client.
> >>
> >> User John Doe authenticates with his credentials and obtains token
> >> pair
> >> *A1R1*
> >>
> >>   1. After A1 expires, app refreshes the token pair to *A2R2 *USING R1
> >> 2. After A2 expires, app refreshes the token pair to *A3R3 USING
> >>      **R1***(ideally it is should use R2 as it is the latest refresh
> >> token)
> >>
> >> In order to achieve this functionality, I was wondering why can't we
> >> use existing last refresh time from User session rather then checking
> >> it in the client session.
> >>
> >> IMHO, adding one more validation in the ValidateToken method in
> >> TokenManager class like this should resolve the problem.
> >>
> >>           // after userSession is determined either for offline token
> >> or online token...
> >>
> >>           if(oldToken.getIssuedAt() <
> userSession.getLastSessionRefresh()) {
> >>               throw new
> >> OAuthErrorException(OAuthErrorException.INVALID_GRANT, "Stale refresh
> >> token - already used");
> >>           }
> >>
> >> Please let me know if you see any pitfalls other than the backward
> >> compatibility for existing keycloak users. I can work with you to
> >> merge this change & test it in the master.
> >>
> >> Best
> >>
> >> Kamal
> >>
> >> ----------------------------------------------------------------------
> >> --
> >>
> >> *From:*Stian Thorgersen <sthorger at redhat.com
> >> <mailto:sthorger at redhat.com>>
> >> *To:* Marek Posolda <mposolda at redhat.com <mailto:mposolda at redhat.com>>
> >> *Cc:* "Jagadevan, Kamal" <kamalakannan.jagadevan at hpe.com
> >> <mailto:kamalakannan.jagadevan at hpe.com>>;
> >> "keycloak-dev at lists.jboss.org <mailto:keycloak-dev at lists.jboss.org>"
> >> <keycloak-dev at lists.jboss.org <mailto:keycloak-dev at lists.jboss.org>>
> >> *Sent:* Wednesday, October 7, 2015 8:38 AM
> >> *Subject:* Re: [keycloak-dev] Same Refresh token can be used multiple
> >> times to obtain access token
> >>
> >> You're right, we'd have to introduce a lastRefresh on ClientSession
> >>
> >> On 7 October 2015 at 14:35, Marek Posolda <mposolda at redhat.com
> >> <mailto:mposolda at redhat.com>> wrote:
> >>
> >>      On 07/10/15 14:23, Stian Thorgersen wrote:
> >>
> >>          We should make this configurable. For those worried about
> >>          security they can enforce new refresh tokens as well as offline
> >>          tokens will replace the old tokens. It would be fairly simply
> to
> >>          implement. If enabled we would only allow refresh token where
> >>          iat is >= the last session refresh time.
> >>
> >>      I was also thinking about this possibility. However if you have 2
> >>      clients and you refresh the token for client1, the refresh token of
> >>      client2 won't be valid as his "iat" will be older. Also SSO login
> >>      currently refreshes lastSessionRefresh on UserSession. However
> maybe
> >>      we can introduce lastSessionRefresh to ClientSession as well?
> >>
> >>      Marek
> >>
> >>
> >>
> >>          I wouldn't make it default behavior for two reasons:
> >>
> >>          * It would break existing clients if they expect to continue
> >>          using the old refresh token
> >>
> >>          * It comes at a performance cost as clients will have to store
> >>          the new refresh tokens and offline tokens each time they
> refresh
> >>          the token
> >>
> >>          * For offline tokens Keycloak would also have to persist the
> >>          last refresh time each time the offline token is refreshed
> >>
> >>          I think we'd need to make it a realm wide configuration option.
> >>
> >>          On 7 October 2015 at 14:12, Marek Posolda <mposolda at redhat.com
> >>          <mailto:mposolda at redhat.com>> wrote:
> >>
> >>              The points are valid and security can be always improved,
> >>              however sometimes improving security makes things
> >>              complicated with the not-so-big advantage... IMO admin
> >>              should always protect the machine to make sure that nobody
> >>              unauthorized has access to refresh tokens. And for the
> >>              transport, HTTPS should be always used. But feel free to
> >>              create JIRA and we will see...
> >>
> >>              When user or client is deleted, all refresh/offline tokens
> >>              will defacto become invalid as well and can't be used
> >>              anymore. You're right that offline token is still valid
> >>              after user logout. User can revoke it manually in account
> >>              management or admin can revoke it in admin console. However
> >>              refresh token is invalid after user logout. All
> >>              refresh/offline tokens for particular client can be revoked
> >>              by admin by set notBefore policy to now, which can be done
> >>              in admin console in "Revocation" tab of particular client.
> >>
> >>              Marek
> >>
> >>
> >>
> >>              On 07/10/15 04:27, Raghuram Prabhala wrote:
> >>
> >>                  Very valid points Mike and even I have similar
> concerns.
> >>                  But please do understand that even if the refresh token
> >>                  is stolen or compromised,it cannot be used by any
> client
> >>                  unless both the client_id and client_secret are also
> >>                  compromised/stolen. But nevertheless, it is a good
> >>                  practice to assume the worst and add in protective
> >>                  measures to minimize the chances.
> >>
> >>                  Marek/Bill/Stian - Even our organization is very
> >>                  particular that such potential security issues be
> >>                  addressed. Can this be taken up? BTW I am not sure if
> >>                  you have an API/End point to invalidate tokens for
> those
> >>                  that are either compromised or must be invalidated as
> >>                  either the user or client is no longer active. If you
> do
> >>                  not have one then it is a good idea to make one
> available.
> >>
> >>                  Thanks,
> >>
> >>                  Raghu
> >>
> >>
> >> ----------------------------------------------------------------------
> >> --
> >>
> >>                  *From:*"Kuznetsov, Mike" <mikhail.kuznetsov at hpe.com>
> >>                  <mailto:mikhail.kuznetsov at hpe.com>
> >>                  *To:* "keycloak-dev at lists.jboss.org"
> >>                  <mailto:keycloak-dev at lists.jboss.org>
> >>                  <keycloak-dev at lists.jboss.org>
> >>                  <mailto:keycloak-dev at lists.jboss.org>
> >>                  *Cc:* "Jagadevan, Kamal"
> >>                  <kamalakannan.jagadevan at hpe.com>
> >>                  <mailto:kamalakannan.jagadevan at hpe.com>
> >>                  *Sent:* Tuesday, October 6, 2015 4:34 PM
> >>                  *Subject:* Re: [keycloak-dev] Same Refresh token can be
> >>                  used multiple times to obtain access token
> >>
> >>                  Hello,
> >>
> >>                  The reason I brought this up is that we are currently
> >>                  working on migrating out authentication from a
> >>                  commercially available product called Ping to Keycloak.
> >>                  We noticed that Ping invalidates the refresh token
> after
> >>                  it is used once, while Keycloak does not.
> >>
> >>                  I and my colleague, Kamal are concerned that by not
> >>                  invalidating the refresh token after first use, we may
> >>                  be opening a security hole. While SSL may protect the
> >>                  token in transit, we can see a scenario where the
> >>                  refresh token would be compromised or stolen from the
> >>                  client itself. In this case, the stolen refresh token
> >>                  could be used to get new access tokens without the
> owner
> >>                  of the client machine knowing.
> >>
> >>                  However, if the behavior was changed so that the
> refresh
> >>                  token could only be used once, then either:
> >>
> >>                  1.If the owner of the client machine would use the
> >>                  refresh token first, then the stolen refresh token
> could
> >>                  not be used
> >>
> >>                  2.If the stolen refresh token would be used first, then
> >>                  the client machine would not be able to use it and the
> >>                  user of that client machine could be alerted that
> >>                  something was wrong. This user could then reset their
> >>                  password or invalidate all of their access and refresh
> >>                  tokens.
> >>
> >>                  Furthermore, we are concerned about this same scenario,
> >>                  but with the offline token. My understanding is that
> the
> >>                  offline token does not expire and that it can't be
> >>                  invalidated by logging out the user or changing the
> >>                  user's password. Have you thought about this scenario?
> >>
> >>                  Thank You,
> >>
> >>
> >>                  *Mikhail Kuznetsov*
> >>
> >>                  Software Engineer
> >>
> >>                  Hewlett Packard Enterprise
> >>
> >>                  *From:*Marek Posolda [mailto:mposolda at redhat.com]
> >>                  *Sent:* Tuesday, October 06, 2015 1:16 PM
> >>                  *To:* Raghu Prabhala
> >>                  *Cc:* Kuznetsov, Mike; keycloak-dev at lists.jboss.org
> >>                  <mailto:keycloak-dev at lists.jboss.org>
> >>                  *Subject:* Re: [keycloak-dev] Same Refresh token can be
> >>                  used multiple times to obtain access token
> >>
> >>                  Hi Raghu,
> >>
> >>                  >From the specs, it looks to me that this is not
> anything mandatory. The paragraph is starting "For example". Feel free to
> create JIRA, but I personally can't promise anything regarding this...
> >>
> >>                  Marek
> >>
> >>
> >>                  On 06/10/15 17:37, Raghu Prabhala wrote:
> >>
> >>                      Hi Marek - section 10.4 of rfc6749 mentions that
> the
> >>                      prior refresh token should be invalidated but
> >>                      retained by the server - to handle compromise of
> >>                      refresh tokens as they are long lived.
> >>
> >>                      Thanks,
> >>
> >>                      Raghu
> >>
> >>                      Sent from my iPhone
> >>
> >>
> >>                      On Oct 6, 2015, at 10:53 AM, Marek Posolda
> >>                      <mposolda at redhat.com <mailto:mposolda at redhat.com>>
> >>                      wrote:
> >>
> >>                          You're right, same refresh token can be used
> >>                          more times. However it is still better to use
> >>                          refresh token R2 in your step 3 instead of
> using
> >>                          old refresh token R1 because R2 has updated
> >>                          timestamp (each token is valid just for 30
> >>                          minutes or so, depends on the configured SSO
> >>                          session idle timeout).
> >>
> >>                          Or are you referring that this is security
> issue
> >>                          and potential possibility to Man in the middle?
> >>                          If you use HTTPS (which is recommended for
> >>                          production environment, and especially if you
> >>                          have unsecured/untrusted networkl), this
> >>                          shouldn't be an issue.
> >>
> >>                          Marek
> >>
> >>                          On 06/10/15 16:34, Kuznetsov, Mike wrote:
> >>
> >>                              Hello,
> >>
> >>                              I noticed that with Keycloak, it seems that
> >>                              refresh tokens are still valid after they
> >>                              are used once. This means that Keycloak
> does
> >>                              *not* invalidate Refresh Tokens after they
> >>                              have been used once.
> >>
> >>                              I am able to successfully execute the
> >>                              following flow:
> >>
> >>                              1.Obtain Access Token (A1) and Refresh
> Token
> >>                              (R1)
> >>
> >>                              2.Use Refresh Token (R1) to obtain new
> >>                              Access Token (A2) and Refresh Token (R2)
> >>
> >>                              3.Use same Refresh Token (R1) again to
> >>                              obtain new Access Token (A3) and Refresh
> >>                              Token (R3)
> >>
> >>                              Can you please tell me if this is the
> >>                              intended functionality?
> >>
> >>                              Thank You,
> >>
> >>
> >>                              *Mikhail Kuznetsov*
> >>
> >>                              Software Engineer
> >>
> >>                              Hewlett Packard Enterprise
> >>
> >>
> >>
> >>
> >> _______________________________________________
> >>
> >>                              keycloak-dev mailing list
> >>
> >>                              keycloak-dev at lists.jboss.org
> >>                              <mailto:keycloak-dev at lists.jboss.org>
> >>
> >>
> >> https://lists.jboss.org/mailman/listinfo/keycloak-dev
> >>
> >>                          _______________________________________________
> >>                          keycloak-dev mailing list
> >>                          keycloak-dev at lists.jboss.org
> >>                          <mailto:keycloak-dev at lists.jboss.org>
> >>
> >> https://lists.jboss.org/mailman/listinfo/keycloak-dev
> >>
> >>                  _______________________________________________
> >>                  keycloak-dev mailing list
> >>                  keycloak-dev at lists.jboss.org
> >>                  <mailto:keycloak-dev at lists.jboss.org>
> >>                  https://lists.jboss.org/mailman/listinfo/keycloak-dev
> >>
> >>
> >>
> >>                  _______________________________________________
> >>
> >>                  keycloak-dev mailing list
> >>
> >>                  keycloak-dev at lists.jboss.org
> >>                  <mailto:keycloak-dev at lists.jboss.org>
> >>
> >>                  https://lists.jboss.org/mailman/listinfo/keycloak-dev
> >>
> >>
> >>              _______________________________________________
> >>              keycloak-dev mailing list
> >>              keycloak-dev at lists.jboss.org
> >>              <mailto:keycloak-dev at lists.jboss.org>
> >>              https://lists.jboss.org/mailman/listinfo/keycloak-dev
> >>
> >> _______________________________________________
> >> keycloak-dev mailing list
> >> keycloak-dev at lists.jboss.org <mailto: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
> >
>
> --
> 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/keycloak-dev/attachments/20151014/0271ba85/attachment-0001.html 


More information about the keycloak-dev mailing list