Hi Alexander,
Thanks a lot for the info and confirmation, that sounds reassuring :)
We're thinking to keep FlatfileAuth as a Plan B option.
But at the moment I think the best would be the "Cert Based Auth"...
So I'd like to narrow down my questions raised earlier, and would really appreciate
your input.
This is an extract from the SCEP specification -
2.1.1.3<http://tools.ietf.org/html/draft-nourse-scep-17#section-2.1.1....;. Requester
Uses Existing CA-Issued or Self-Signed Certificates
In this protocol, the communication between the requester and the certificate authority is
secured by using PKCS#7 [
RFC2315<http://tools.ietf.org/html/rfc2315>] as the
messaging protocol (see Section
4<http://tools.ietf.org/html/draft-nourse-scep-17#section-4>). PKCS#7
RFC2315<http://tools.ietf.org/html/rfc2315>], however, is a data format which
assumes the communicating entities already possess the peer's certificates and
requires both parties use the issuer names and issuer assigned certificate serial numbers
to identify the certificate in order to verify the signature and decrypt the message.
* If the requesting system already has a certificate issued by the CA, that
certificate SHOULD be presented as credentials for the renewal of that certificate if the
CA supports the "Renewal" capability and the CA policy permits the
certificate to be renewed.
* If the requesting system has no certificate issued by the CA, but has
credentials from a different CA, that certificate MAY be presented as credentials instead
of a self-signed certificate.
Policy settings on the SCEP server will determine if the request can be accepted or not.
* If the requester does not have an appropriate existing certificate, then a self
signed certificate must be used instead. The self signed certificate MUST use the same
subjectName as in the pkcs10 Request.
One of those certificates should be used to sign the overall PKCS#7 envelope sent by
clients in "PKCSReq" message to the CA. My hunch is that iOS devices use
identity certificates issued by Apple for signing SCEP messages (i.e. PKCS#7).
Could anyone confirm that the second and third bullet points (i.e. credentials from a
different CA & self signed certificate) are supported in Dogtag?
If yes, could you also clarify one more thing please?
* Does the CA compare the CN in the signing cert with the CN in the PKCS#10 ?
With thanks,
Oleg
From: Alexander Jung [mailto:alexander.w.jung@gmail.com]
Sent: 21 August 2013 20:32
To: Oleg Antonenko
Cc: Andrew Wnuk; pki-users(a)redhat.com
Subject: Re: [Pki-users] Using SCEP
Hi,
we gathered some experience using scep and dogtag. To be a little more precise we are
issuing certs to cisco routers by the thousands (the whole sales force needs these...)
The current implementation works, but leaves still some space for improvement :-)
Your client initially needs a password that must be known to the ca in the
flatfileauth-file used for your scep profile. The CA has a simple (more example)
application to request such a password, we tied it to our order system to further
authenticate those requests.
When your CA certificate (or the certificate you are using for scep) sits in a HSM,
you'll need quite an extension for the existing code, as the current code will not be
able to decrypt the requests (in this case due to ciscos error - but we have to serve our
clients...)
The flatfileauth module still has a longstanding bug (from the iplanet days of the dogtag
code), that prevents it to work with other tag-names than the default ones, easy to fix,
but hair tearing when you debug it. (See
https://www.redhat.com/archives/pki-devel/2009-February/msg00000.html for details)
The cisco client still complains on some operations not being implemented, but works after
those modifications
Yours,
Alexander