CNN.com on Remailers

Bill Stewart bill.stewart at pobox.com
Mon Dec 31 21:10:42 PST 2001


Depending on the recipient, you might or might not be encrypting the message.
But the important security you're protecting is the connection between
the sender and the recipient.  Depending on the application,
the sender may be trying to prevent the recipient from knowing his address,
or the two of them may be trying to prevent outside eavesdroppers from
knowing that they're communicating, but in either case,
you only have that security if you can trust the remailer system
not to divulge the relationships.  If you let the sender's system
do all the encryption, then the entire chain needs to be compromised
for the connection to be revealed; otherwise any remailer along the way
can rat him out.  That doesn't mean that you can't gain some security
by remailers in the middle also adding hops on their own,
but the sender can't depend on that (at least unless he trusts
the remailers that are doing that), and of course you risk routing loops,
which are ugly things to detect in connection-obfuscating environments.


At 10:02 AM 12/29/2001 -0600, Jim Choate wrote:
>On Sat, 29 Dec 2001, Bill Stewart wrote:
>
> > > At 09:01 PM 12/17/2001 -0600, Jim Choate wrote:
>
> > The only way to get security is for the originator to do the encryption -
> > otherwise, if ANY remailer in the chain is compromised,
>
>Actually this isn't the 'only' way. ALL (!!!) that is required to keep the
>security of the email traffic is that it is source encrypted for the
>destination; it's gibberish to all middle-men. What the remailer chain
>does is break the causal connectivity, it provides plausible deniability.
>....
> > the Bad Guys can read the message.
>
>At no point can anyone other than the recipient 'read the message', unless
>it was sent in the 'clear' in the first place (silly thing to do).
>
> > If the originator does the crypto,
> > then EVERY remailer in the chain has to be compromised to break it.
>
>ROTFLMAO. ONLY(!!!) if the source didn't destination encrypt to begin
>with. A critical step you seem to not quite 'get'.








More information about the cypherpunks-legacy mailing list