[Lake] EDHOC exporter use patterns

["Sipos, Brian J." <[e-mail]>]

All,

I'm looking into detail design for an application embedding of EDHOC which
needs to derive multiple shared secrets across a long time period. It would
be logically acceptable for this application to use a single EDHOC_Exporter
[1] label with unique contexts for each shared secret. It's not clear to me
whether there is an expectation that the `PRK_exporter` state would be
preserved for as long as an application might need to use it, or if
applications are expected to do any exporting immediately as the EDHOC
messages are being processed and not at all afterward. Is there some detail
that I am missing from current RFC 9528?

 

For example, the OSCORE use of EDHOC derives a single immediate set of
secrets (the "master secret" and "master salt") and uses those for later
derivations of OSCORE-specific shared secrets. Since I don't know of any
other current users of EDHOC it's difficult to determine if this pattern
must be followed or if there is allowance for other patterns of exporter
use.

 

While this might seem like an implementation matter, I would prefer to not
require behavior from an EDHOC API that does things in unexpected or
invasive ways. A two-phase derivation pattern will get the same resulting
need but might be more conforming to EDHOC use expectations.

 

Thanks for any feedback,

Brian S.