On-line certification in browsers
=========================

Keygen is a poorly documented Netscape "relic" that do not support things like
- Key export policies
- PIN-code policies
- Key container PoP
as well as not offering an integrated mode for the entire key-gen to
cert-download process.  The latter becomes particularly ugly
when different keys are deployed for encryption, authentication
and digital signatures.

Xenroll is Microsoft's proprietary scheme which is considerably more
advanced than keygen but still depends on MS-only stuff.

Due to this a number of larger CAs have excluded both of these schemes
in favor for proprietary but still cross-browser-usable schemes based on
Java applets.

Question: Is there anybody on this that would be interested in creating a
new scheme that could gradually replace the dated schemes above?

You may want to look at a recent counterpart for OTP tokens:

http://www.rsasecurity.com/rsalabs/node.asp?id=2817

Compared to PKCS #10 and CRMF, CT-KIP is:
1. XML based
2. Multi pass

That is, revising existing RFCs is not likely to be a good idea since
on-line certification needs multi-pass protocols in order to support
users with limited, or no knowledge of PKI.  Certificate renewal
also requires multi-pass including an atomic replacement operation.

A difference with on-line certification schemes compared to for example
the creation of web-server certificates is that in the former case the CA
is usually the originator of all vital information including key length, subject
DN etc.  This is principally about the opposite of PKCS #10 and CRMF.

So IMHO a new CSR scheme should [at least] allow the CA to define
- key length and type
- PIN/passphrase policy
- key export policy
- All but the public key of the EE certificate(s) to be generated

As well as optionally requiring HW key container PoP & ID.

The goal must be to eliminate as much as possible, any requirements on
the user regarding answering questions that they would never have to
bother with if they got a token distributed in a physical form, while still
providing for a secure process.

It should also support TPMs which are just around the corner.

Anders Rundgren
_______________________________________________
mozilla-crypto mailing list
mozilla-crypto@mozilla.org
http://mail.mozilla.org/listinfo/mozilla-crypto

Reply via email to