ietf-dkim
[Top] [All Lists]

[ietf-dkim] bot-nets

2008-06-26 12:32:21
In the security considerations section, there were two paragraphs  
below DNS Attack added.  This was intended to be within a section  
called Bot-net Attack.

Why this section is important might be illustrated by with the  
following data.

A well respected email provider has a slight problem.  They send  
approximately ten billion emails a day, or about one hundred thousand  
each second from 2 million cached daily inquires from customers  
assumed to represent about 1/5000th of overall traffic.   It is clear  
that they understand there is a problem from their arrangement of  
outbound servers.  They have three tiers earning significantly  
different, but uniform ratings.  The first class tier sends little if  
any spam, the second class sends a small percentage of spam mixed  
together with about 2 billion messages, and the third class sends  
about 350 million a day with a high percentage being spam.  This  
provider publishes SPF records, and, in the past, asked why this space  
had not been white-listed.

It should not be difficult to understand the danger this provider's  
behaviour might represent.  As it would be for DKIM signatures and  
 From headers, the PRA is not restricted.  In the case of Sender-ID, a  
recipient of their message might be convinced to process perhaps nine  
macro expanded SPF records evaluating a PRA targeting some victim, and  
yet the message could still be authorized when the bad-actor finally  
includes the provider's highly respected SPF record.  What a concept!

I was prevented by several MAAWG members from making a presentation  
pointing out the dangers associated with Sender-ID then being touted  
by a new board member and soon highly promoted as an industry  
solution.  After all, this was a feature promised to enrich those  
supporting it.  There were even direct payments made to implementers.   
(The problem that SPF sans Sender-ID hoped to solve is more safely  
handled by adoption of RFC3834.  A reputation service could help make  
that happen.)

DKIM is intended to help solve the phishing problem.  IMHO, it can.   
However, the DKIM WG must not ignore DDoS issues.  It does not matter  
how many major email providers insist their mail does not represent a  
concern.  If ADSP were to limit how the i= parameter within the DKIM  
signature can be used, or insist that the "on-behalf-of" feature can  
be obtained by using multiple signatures, or by having the identity  
introduced through some undefined method, then clearly this WG has  
failed to appreciate the the danger as well.  Although admitting there  
is a problem may draw the interest regulators, where little good ever  
comes, by ensuring DKIM is able to help mitigate the bot-net issue, it  
can better head off the dangers associated with any eventual regulation.

-Doug
_______________________________________________
NOTE WELL: This list operates according to 
http://mipassoc.org/dkim/ietf-list-rules.html