Re: [Strint-attendees] Do we need a breakthrough in key management first?

[Phillip Hallam-Baker <[e-mail]>]

I think we need to fix key management but what it takes is a different
perspective. For the past twenty five years everyone has been looking for
the solution to the key management problem. Well perhaps that is because
there isn't one. Perhaps there are many problems that require different
solutions.

This presentation argues the case rather more thoroughly:

https://www.youtube.com/watch?v=PBFnBpWkK8M


The point is that PGP and PKIX both work. But they solve different
problems. So people who start of their work on key management with 'first
lets kill all the lawyers' aren't helping.

Rather than looking for the perfect key management scheme we should build a
framework that separates as much common 'plumbing' that can be shared from
the trust infrastructure problems where multiple approaches are needed. We
don't need separate S/MIME and PGP/MIME message formats but we do need
direct trust, trusted third party trust and peer to peer trust models
because each solves different needs.

Some parts of key management could be taken off the table with shared
standards. Management of private keys is too much of a pain. It takes a lot
of unnecessary effort to generate keypairs and move them between machines.
None of the existing systems has a viable key recovery scheme for end users.

Further if we combine CT concepts with trusted third party trust and peer
to peer trust models we get an improvement on either alone.

But we can only see an improvement if we have a basis for measurement.
Which is why I believe that we need to use the time based work factor as
the metric for comparing trust models. We compare different algorithms by
the computational cost of breaking them. We should adopt the same rigor to
trust models. I am rather fed up of hand wavy arguments about how there are
'billions' of CAs and therefore the new model someone has just pulled out
of whatever has to be a 'better' solution. I exaggerate but only by a
little.


It is also probably a mistake to base argument for protecting pervasive
security to heavily on examples drawn from dissident movements who know
that they are under observation and face real risks if caught. Dissident
movements have no problem with using PGP but that is because they have
incentive to learn and make effort. Expecting the other 2 billion users to
adopt the same technology is rather optimistic.