ietf-mxcomp
[Top] [All Lists]

Granularity of Reputation (was Re: Against Extensibility in MARID Records

2004-06-20 18:42:35

On 6/18/04 9:05 PM, Hallam-Baker, Phillip sent forth electrons to convey:

I think that there is a good case for tying reputation to a domain.

But not having actually solved the problem yet (the spam is still
with us), I am not going to argue against tying accreditation or
reputation to a finer granularity.

I think it is very likely that marketing(_at_)anybank(_dot_)com is going to
be in a different reputation category than 
jane(_dot_)doe(_at_)anybank(_dot_)com(_dot_)

I was thinking more about granularity. I think changing granularity from the domain level essentially introduces several security flaws into the system.

If I were running a reputation service, I'd be very reluctant to make it possible for the above emails to have separate reputations, or to allow marketing.anybank.dom and billing.anybank.com to have separate reputations. I'd end up playing whackamole if I wasn't extremely careful; email addresses and subdomains are free. I'd *have to* charge for such entries, even if I wanted to run a free service, for it not to be fundmentally broken.

Other whackamole security flaw secenarios this would enable: spam run begins, authorized by spammer.dom. SPF record of spammer.dom is changed to redirect from one throwaway, or ?all domain to another, as soon as each is blacklisted. Makes much more sense to blacklist spammer.dom.

In other words, (in SPF terms), reputation MUST be tied to <responsible-sender>. It may in addition be tied to <current-domain>.


<Prev in Thread] Current Thread [Next in Thread>
  • Granularity of Reputation (was Re: Against Extensibility in MARID Records, Matthew Elvey <=