Re: [Iot-onboarding] [Anima] Device Certificate Deployment Automation with ACME using BRSKI

[Kent Watsen <[e-mail]>]

Hi Michael,


>> Skimming quickly, I see now the direction to go to a cloud registrar to
>> be redirected to a local registrar.  I feel compelled to point out that
>> this is exactly what SZTP (RFC 8572) does, or at least, supports.
>> Actually, as a more general statement, it was originally said that the
>> two WG's approaches were different, but they are now becoming more and
>> more alike.  Well, since SZTP is published already, it's more like the
>> ANIMA approach is becoming more like SZTP, albeit seemingly with more
>> complexity.
> 
> SZTP does not directly result in the same mutually authenticated TLS (EST)
> connection to a Registrar that BRSKI does.  This was not an explicit SZTP goal.

It's true that EST isn't used, but there is a mutually-authenticated TLS connection to the SZTP bootstrap server, which is effectively the same as BRSKI's Registrar.   


> What it could do though, is to provide a source of (secure) bootstrap
> configuration in which such a connection could be initiated.  In my mind,
> I see a fragment of signed JUNOS, IOS (CISCO, not Apple), etc. CLI
> configuration that initiates the connection. Such a thing would not be
> standardized.
> But, I would think that this is exactly the kind of thing that could
> be encoded in YANG.

Indeed, this what RFC 8572 describes (in section 5.6 and also in C.3, item #3) and the primary motivation behind https://tools.ietf.org/html/draft-ietf-netconf-restconf-client-server <https://tools.ietf.org/html/draft-ietf-netconf-restconf-client-server> and https://tools.ietf.org/html/draft-ietf-netconf-netconf-client-server <https://tools.ietf.org/html/draft-ietf-netconf-netconf-client-server>.



> My understanding of https://tools.ietf.org/html/draft-ietf-netconf-keystore
> is that it would provided for Registrar initiated (PUSH) updating of device
> certificates, but would not provide a way for a device to initiate (PULL) to
> a securely identified EST server.

Sounds correct, but clarifying:

1) the current keystore model is all about enabling a controller/NMS application to configure/set/push keys and associated end-entity certificates to a device.

2) there is a suggestion that the keystore model could/should be extended to support ACME (or similar), in which case one might claim that the device had "pulled" an end-entity certificate from a "securely identified server" (dropping the "EST" part).

3) the truststore model (draft-ietf-netconf-trust-anchors) can be used by a controller/NMS application to configure/set/push trust-anchor certs used, e.g., to verify a remote server's end-entity certificate.


Cheers,
Kent