[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: REMAIL: X-TTL and X-Discard



Hal wrote: 

> 

> What is needed to make X-TTL useful is for the remailer to choose another
> remailer as its destination, and ideally to encrypt the message before
> sending it.  This way X-TTL can be used to insert a random remailer path of
> n hops in the middle of a sender-constructed remailing path.  This leads to
> a system where the remailer decrypts an incoming message, reads the X-TTL
> value, decrements it, re-encrypts the message for the next remailer in the
> chain, and sends it.  The X-TTL value is never exposed to outsiders.
> 

> At one point I wrote a modification to my remailer to cause it to
> encrypt any message which it sent to another remailer which supported
> PGP.  But I decided that this didn't really help security enough to
> be worthwhile.  It would be much better to encourage users to encrypt
> their messages themselves in a nested fashion so that no remailer sees
> any more information than the bare minimum necessary.

Rolling your own encryption wrapper for the remailer chain you're sending  
through is a Good Thing, but your modification would be useful if you think of  
the cypherpunk remailer network as a "back end" for an anonymous/pseudonymous  
server like Julf's.

Ideally, a pseudonym server will only keep an encrypted remailer chain for a  
user's return address (along with the unencrypted adress of the first remailer  
on the chain).  The nymserver _doesn't_know_ what remailers are in the chain,  
so it can't encrypt the message with each of their public keys.  But if the  
server can include a header line inside the encryption envelope that tells the  
remailer to encrypt with the next remailer's key, we can be sure that an  
adversary is still unable to match up incoming and outgoing messages.

Setting up a pseudonym server with this kind of encrypted return address is  
good, of course, if you're worried about its database being seized.  Without  
the cooperation of each remailer in the chain, the database doesn't give an  
adversary anything useful.

And since now we've got TTL as another use for a next-step-encryption feature  
in the remailers...  I'd better go get those remailer scripts and a UUCP feed  
for my new Linux box.

Joe