ietf-822
[Top] [All Lists]

Re: Sender authentication?

2004-02-09 00:23:10

Nick Ing-Simmons wrote:

I hope this isn't re-hashing old ground but an idea has occured to me
that would, I think, help with the current deluge of mail from worms that forge From: addresses.

The idea is to "sign" the From: and Message-ID: pair using a public key scheme, and add the signature as a new header.

How do you want to check signature validity for John Doe ?
Do you want to use it only for "well known friends" ?

The goal is to allow recipient to have some confidence that mail is really from the sender without collecting the whole mail, but just by examining the headers. (Signed mail is all very well but sucking 5M of spam body to allow me to check it isn't helping with my bandwidth problem.)

Only POP offers an option to download headers without body. Well behaving SMTP servers can not accept headers but refuse to accept body.
Virus on A person host can "reuse" signature of B->A message to fake B->C 
message.

(I currently discard any mail which has its Message-ID inserted by my ISP - this is wonderfully successful for an ad. hoc. hack e.g. it managed to discard 80% of the 10000 emails I received in one day of recent MyDoom peak. Snag is that there are presumably false discards - including mails from my ISP!)

Viruses get smarter every outbreak.

The idea is that the unique Message-Id is a "challenge" for the sender's
key pair. As many tools also discard mails with duplicate ids then forger cannot just re-use signature headers from a previous mail.

Issues I can see are where public key comes from (key server or embedded in the mail) and possible size of key and signature data.

Is this worth discussing further?

The idea is good BUT IMHO signing Sender, Recipient, Message-ID, Date makes more sense.
* "Date:" signing would prohibit "reuse" of signatures.
* Recipient signing is tricky vide signing BCC: recipients

--
Andrzej [en:Andrew] Adam Filip http://anfi.freeshell.org backup: 
anfi(_at_)xl(_dot_)wp(_dot_)pl


<Prev in Thread] Current Thread [Next in Thread>