ietf-asrg
[Top] [All Lists]

[Asrg] 0. General - Thoughts on testing anti-spam software

2004-03-08 07:42:36
[this email rambles a lot. apologies]

This is perhaps a little bit off-topic, so I'm as happy to recieve
replies off-list. This month, the magazine I work for
(http://www.virusbtn.com) is conducting its first test of anti-spam
software. The current plan is to kick off with testing three open-source
lexical analysers - CRM114, DSPAM, and Bogofilter. I've been talking to
some of the big commercial anti-spam people about testing, and how I
plan to test these, and a recurring idea is that while they feel these
solutions will perform exceptionally well for one user, they just don't
scale for multi-user installations...

I've had a think about this. 

The main objection seems to be the fact that you need to maintain
keyword lists for every user. This is seen to be a big problem,
although, looking at my SpamAssassin config directory, my 'keyword
dictionaries' are about 6MB combined. Even with 1,000 users, we're not
talking about a great deal of data retention here... 

Has anyone installed a bayesian/statistical based filtering solution for
a big user-base, and wants to share their experiences of how effective
it's been? How much space it's taking up? I'd be really curious to hear.

We have a number of corpuses we're testing against, put together in
different ways. The review will probably explain each corpus, and what
its strengths and weaknesses are. We're going to be keeping results from
different corpuses different - we won't be testing a product against a
corpus when it's been trained on a different corpus.

Any thoughts, on general anti-spam software testing, or on any points
raised here, welcomed.

Thanks

+Pete

-- 
Do not be too moral. You may cheat yourself out of much life. Aim above
morality. Be not simply good; be good for something.
 -- Henry David Thoreau

_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg



<Prev in Thread] Current Thread [Next in Thread>
  • [Asrg] 0. General - Thoughts on testing anti-spam software, Peter Sergeant <=