'Today' considers data retention and IMP

[Chris Edwards]

On Mon, 12 Jan 2009, Igor Mozolevsky wrote:

| 2009/1/11 Chris Edwards:
| 
| > I see an increasing number of mail systems, including those operated by
| > various UK Universities, that can now *only* be accessed by their users
| > via the TLS versions of IMAP / SMTP / webmail.
| 
| But these encrypted channels only go between the end user and the
| server. SMTP-to-SMTP transactions are still done mainly in plaintext.

Yep.  Even with the user<->server IMAP+SMTP traffic encrypted, black boxes 
sitting on backbone links would typically see server<->server SMTP traffic 
in the clear.

This raises in interesting point...

A fair number of mail-servers happily send and/or receive encrypted SMTP 
when talking to remote mail *servers* with similar capability.  Unlike the 
client<->server case, which normally involves a proper certificate check, 
the server<->server case usually does not.  So we only get opportunistic 
encryption, which defeats passive sniffing attacks, but is vulnerable to 
active middle-person attacks.

So, I wonder whether the IMP black boxes will perform the active attacks 
needed to access server<->server opportunisticly encrypted email ?  This 
would seem to go against the traditional intelligence mantra of entirely 
passive listening devices...