spf-discuss
[Top] [All Lists]

Re: using social networks to evaluate senders

2005-01-21 10:14:24
* Meng Weng Wong <mengwong(_at_)dumbo(_dot_)pobox(_dot_)com> [2005-01-21 
12:03-0500]
CAMBRIDGE, MA --- The MIT Spam Conference is under way.
Most of the presentations follow the content filtering
school of thought.  I, personally, feel that content
filtering has reached the point of diminishing returns.
Sender authentication, which SPF pioneered, allows us to
ask a different question: "are you a stranger?"

We already know that addressbooks are only the most obvious
way to answer that question.  Social networking offers a
richer answer: if you're three degrees of separation away
from me, that's probably good enough.

P. Oscar Boykin today announced some interesting results in
this space.  It was previously covered on Slashdot:

 http://slashdot.org/articles/04/02/19/1542248.shtml

Here is the abstract from the talk:

 http://www.spamconference.org/abstracts.html#Boykin

Loaf is another example of work in this area:
 http://loaf.cantbedone.org/
 http://dumbo.pobox.com/~mengwong/tmp/loaf-diagrams/mouseovered.html

+1 on this general approach. I think there'll come a point where 
submissions to wikis, blogs etc. use some variant of this. I had a 
(cruder than LOAF) experimental FOAF whitelist sharing effort a 
couple years ago, but didn't push it since people were concerned 
about the hashed mailboxes being brute-force reversable at least in 
case of common name/domain combos. LOAF probably addresses those 
concerns.

If folks have proposals for updates to the FOAF spec[1] to make 
these kinds of apps easier, I'd be happy to hear them 
(please cc: the FOAF project list, 
rdfweb-dev(_at_)vapours(_dot_)rdfweb(_dot_)org too).

cheers,

Dan

[1] http://xmlns.com/foaf/0.1/


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