I can't see much point in applying and removing PGP and/or S/MIME at the
start and end of the TLS tunnel. If you are worried about the encryption
strength then fix that. If you are worried about the downgrade attack then
hook a policy layer in.

Message layer security will always cover less than transport on an
individual hop because the message layer encryption can't cover the routing
data.


The advantage of going to message layer security is when the start and/or
end point might be different.

So for example, right now almost nobody is equipped to accept encrypted
email as conveniently as unencrypted. I can only receive encrypted email on
one machine. I don't enable all my machines because it is a hassle and the
keys expire etc.

Sending me a message TLS encrypted, I can only decrypt at the inbound mail
server. A PGP or S/MIME message could be decrypted at the inbound mail
server OR at the mail client depending on what gives the best balance of
performance / security / spam control / compliance and convenience.


The other advantage to using message layer security is that it is possible
to force use of encryption. So for example, let us imagine that I have an
outbound mail server that knows how to resolve key identifiers to public
keys using some protocol (Vcard, WebFinger, WKS, wev). I can poke the
outbound mail server to automatically encrypt messages if the email address
has a particular escape code in it. At the moment I am using a question
mark.


Let us imagine I want to send an email message to Jacob that I want to be
sent encrypted or not at all. I would use the address
[email protected] which
tells my outbound mail server 'use whatever resolution services are
available to find a key and if a trustworthy key can be found use it to
send the message, otherwise report a delivery failure.


If I really want to be sure that the key is correct then I would explicitly
specify the key fingerprint:

228F-AD20-3DE9-AE7D-84E2-5265-CF9A-6F91-4193-A197?ja...@appelbaum.net


The nice thing about this approach is that I can use it with all my
unmodified mail accounts and mail clients. All that I do is to redirect the
outbound mail service through my trusted outbound mail gateway (which is on
127.0.0.1 on most of my machines). I could even send messages through my
Gmail account (but would have to use a non Webmail client to compose and
send).

Receiving mail requires me to either use a mail client with my S/MIME cert
loaded or an S/MIME viewer for WebMail.


At the moment this is a crazy hybrid of S/MIME and PGP approaches. But the
fact is that S/MIME has pervasive deployment while PGP has a userbase. Plug
ins are a hack for testing purposes only, any security scheme that depends
on typical users deploying a plug in is going to fail. It is painful enough
dealing with the vagaries of the platform and application update mechanisms
changing stuff on a daily basis without a layer of plug ins to cope with on
top.
_______________________________________________
perpass mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/perpass

Reply via email to