Re: [Webpush] CALL FOR CONSENSUS: VAPID cut-and-paste protection

(1)  we did discuss using mutual TLS for webpush, as one of the early
proposals, which would mitigate this
issue.  However it was felt that support for mutual TLS would be tricky in
various environments, limiting the adoption
both client and server side.  With the work going on SPIFFE and other
projects (gRPC, k8s) -  this may change and
 IMHO a future revision of webpush could add it.

On VAPID - since it is based on JWT, we have the same limitations as most
other uses of JWT.
I believe Oauth1 was mentioned in some discussions, but more as a thing to
avoid.
Also discussed the fact that bearer JWTs tokens with audience and similar
expiration are widely used
 for protecting more valuable user accounts, and in the case of webpush we
also have the authenticator
 and the (secret) public key.

If VAPID is used for other purposes besides webpush - I believe the format
is extensible enough to
support additional elements to avoid reply.

Costin

On Thu, Aug 17, 2017 at 7:58 PM, Phil Sorber <sorber@apache.org> wrote:

> This is a call for consensus for an issue relating to
> draft-ietf-webpush-vapid, which is currently in IESG evaluation. Interested
> participants should respond no later than Friday, September 1st 2017.
>
> During its initial review, one of the Security Area Directors expressed
> concerns regarding the cryptographic properties of the JWT:
>
> https://mailarchive.ietf.org/arch/msg/webpush/HYW9NcUioQo5X2Np-d2hjCzB1Fo
>
> Specifically: as implemented, the JWT is merely a bearer token. While the
> DISCUSS provides a thumbnail sketch of how this could be mitigated, the
> crux of the issue isn’t the specifics of the implementation, but whether
> the WG had considered other, more cryptographically secure approaches.
>
> Although participants are free to respond in any way they choose, the most
> useful input would be of one of the following three forms:
>
>
>    1.
>
>    I believe the working group has already discussed adding such a
>    mechanism and rejected it (with citation to an email discussion or minutes
>    reflecting such discussion).
>
>    2.
>
>    I do not think the working group has discussed the issue before,
>    however I am opposed to changing the mechanism prior to publication
>    because...
>
>    3.
>
>    I do not think the working group has discussed the issue before, and
>    would support bringing the document back to the working group for the
>    purpose of mitigating copy-and-paste attacks.
>
>
> Thank you.
>
> _______________________________________________
> Webpush mailing list
> Webpush@ietf.org
> https://www.ietf.org/mailman/listinfo/webpush
>
>