Re: [Model-t] Considerations for modern COMSEC protocol design

[Stephen Farrell <[e-mail]>]

Hiya,

On 17/02/2020 21:15, Jim Fenton wrote:
> I'm somewhat less of a fan of opportunistic security; 

Yeah, I think it's fair to say that very reasonable people
do have differing opinions about OS. (That might mean that
Ekr and Chris were wise to say less about it than unwise me
would naturally emit:-)

> before deciding
> that the capabilities required for an active attack are harder to mount,
> you need to consider the threat model. Consider that there are several
> countries that have been launching downgrade attacks against SMTP
> STARTTLS on a substantial fraction of their outgoing traffic [1]. And
> that doesn't even consider more targeted attacks against specific clients.

Sure, [1] is a fine paper and yes such attacks have been
mounted.

However, I don't think there was ever a point in time when
it'd have been possible to seriously argue that *everyone*
MUST turn on TLS authentication in STARTTLS - that just
isn't going to happen all at once at all MTAs. Whereas,
it was entirely possible to encourage everyone to at least
turn on TLS, and enough people have now done that that I
think MTA-STS is now quite credible. (Noting that MTA-STS
comes with the testing and reporting stuff which I'd argue
is also maybe needed in that case.)

Cheers,
S.