Re: comments on web crypto API: Restricting access to private keys to specific servers [4/6]

On Thu, May 23, 2013 at 1:43 AM, Nikos Mavrogiannopoulos
<nikos.mavrogiannopoulos@esat.kuleuven.be> wrote:
> Instead of restricting keys to a specific set of hosts, we propose a
> cryptographic binding of keys to a certain public key.

Thank you for your feedback.

Our goal is NOT to invent "yet another web security" model.

What you describe can be accomplished independently by a web origin by
using key pinning. There is no need to incorporate that normatively
into this API.

There is no effective value added by "re-inventing" HTTP Public Key
Pinning, and significant disadvantage for developers, users, and
implementors.

Can you please explain why HPKP would be insufficient in an
origin-based security model?

>
> Embed a server's public key (S) in the javascript and associate any possibly
> generated private keys with that key. Those keys will be accessible to any
> server that has the (S) key. This of course requires the server to prove the
> possession of the private key that corresponds to (S) to the client. That
> can be done by a signature on some nonce provided by the client (e.g., in
> the HTTP headers). The servers that possess this key should be able to
> enumerate, delete and use the keys generated by them.
>
> On server key compromise a process to update the server key should be
> allowed (e.g. using similar ideas from tack or pinning).
>
> http://tools.ietf.org/html/draft-perrin-tls-tack-02
> http://tools.ietf.org/html/draft-ietf-websec-key-pinning-04
>
>
>

Received on Thursday, 23 May 2013 17:00:12 UTC