On Thu, Jun 20, 2013 at 5:21 PM, Kris Borchers <kris@redhat.com> wrote:
If it's just to meet the new requirement of the server, that's fine.

somewhat, yeah
 
As long as we are aware that there is no security gain from this in JS land and that we make our users aware of this as well.

Currently the endpoint for device registration is "somewhat" secured. Of course JS is odd here.... but even if one knows the variantID and password,
he could not mess around with the Android/iOS variants (they have other ID/secrets).


Perhaps... for SimplePush we can just have a "100% unsecured" endpoint? (instead of doing BASIC with (variantID:secret).


Question: Is it possible... to check with a SimplePush server if a "channelID" is valid (for the simple-push server) ?
So that we can check that channelID "foo" does NOT exist on the SimplePushServer, so we ignore the "registration attempt"


Regarding variantID:secret, I guess not even something like "CryptoJS" can help much here... since it has to parse the "input"


-Matthias







 

On Jun 20, 2013, at 10:18 AM, Kris Borchers <kris@redhat.com> wrote:

So I'm trying to figure out what we gain from this … now instead of putting your variantID in the JS, you are putting the variantID and a secret in the JS. I don't see any security gain here so trying to figure out what we gain.

On Jun 20, 2013, at 10:12 AM, Matthias Wessendorf <matzew@apache.org> wrote:

Hi,

with the use of this helper, it is "safe" (I think) to use the window.btoa function(see details), to perform a (simple) Base64 encoding.

Base64 encoding is required, since the "Device Registration" HTTP REST endpoint now uses HTTP_Basic (for details see the matching thread).

Currently we perform this code for "channel registration":

$.ajax({
  contentType: "application/json",
  dataType: "json",
  type: "POST",
  url: url,
  headers: {
    "ag-mobile-variant": variantID
  },
  data: JSON.stringify({
    category: messageType,
    deviceToken: endpoint.channelID,
    clientIdentifier: alias
  })
});

As mentioned on the "Security thread", the variantID is no longer a header, it is part of the HTTP_Basic auth process.

This is a (local) JavaScript change that I did. It works fine so far:

$.ajax({
  contentType: "application/json",
  dataType: "json",
  type: "POST",
  crossDomain: true,
  url: url,
  headers: {
    "Authorization": "Basic " + window.btoa(variantID + ":" + secret)
  },
  data: JSON.stringify({
    category: messageType,
    deviceToken: endpoint.channelID,
    alias: alias     ///// NOTE:: the key has changed..........
  })
});

The important thing: we add the "Authorization": "Basic " header and using the mentionedwindow.btoa() function for the actual encoding.

The same applies for the DELETE (unregistration).

Any thoughts? Otherwise, I'd send a PR.

Ah.... the dependency agains the Base64.js polyfill library would/should be included in our "grunt" build for "distribution", or would it be "just" declared (yeah, that's details but asking for curiousity)


--
Matthias Wessendorf

blog: http://matthiaswessendorf.wordpress.com/
sessions: http://www.slideshare.net/mwessendorf
twitter: http://twitter.com/mwessendorf
_______________________________________________
aerogear-dev mailing list
aerogear-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev

_______________________________________________
aerogear-dev mailing list
aerogear-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev


_______________________________________________
aerogear-dev mailing list
aerogear-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev



--
Matthias Wessendorf

blog: http://matthiaswessendorf.wordpress.com/
sessions: http://www.slideshare.net/mwessendorf
twitter: http://twitter.com/mwessendorf