From: danisch@ira.uka.de (Hadmut Danisch)
Date: Tue, 8 Feb 94 11:16:50 PST
To: cypherpunks@toad.com
Subject: Insecurity of anonymous remailers
Message-ID: <9402081913.AA00297@deathstar.iaks.ira.uka.de>
MIME-Version: 1.0
Content-Type: text/plain



> From owner-cypherpunks@toad.com Sat Feb  5 21:55:58 1994
> Date: Sat, 5 Feb 1994 15:21:18 -0500 (EST)
> From:  MatthewJ Ghio <mg5n+@andrew.cmu.edu>
> To: Cypherpunks Mailing List <cypherpunks@toad.com>
> Subject: Info on anonymous remailers
> Content-Length: 1429
> 

Matthew J Ghio wrote:

> I am pleased to report on the performance of our two newest remailers,
> qwerty@netcom.com and nate@vis.colostate.edu.  Both remailers had a very
> good response time.

Is it really a good idea to make anonymous remailers work so fast?
Everyone who can analyze the traffic of anonymous remailers and
can read the from/to header lines, the message size and the 
transfer dates immediately knows who sent mail to whom. 

The make the remailers more safe, I would suggest to modify the
software:

Every message must be sliced into pieces of standardized size
( e.g. 8K or 500 lines ). It should be possible to send the
pieces over different channels. Every server should collect 
pieces and keep them a random time. Perhaps they could be
collected until 10MByte of pieces are collected and then they
can be sent out all. Every piece could contain a time limit
which may not be exceeded by the servers. Higher limit means
more confusion for a spy trying to follow this packet. The
packets should be recyphered by every hop. They could have
a random tail who's length is changed by a hop. A traffic
spy could not find out which input and which output packets
correspond.

Hadmut