[Spambayes] Spam Bayes use in a corporation

Ryan Malayter rmalayter at bai.org
Thu May 20 00:55:45 EDT 2004

[Tony Meyer]
> I know almost nothing about IHateSpam, but if I understand 
> correctly, it's basically just white listing, yes?  (i.e. it 
> filters out mail from people you don't know).  This is a 
> fairly weak approach, but popular :)  

Actually, we use iHateSpam for Exchange Server because it enables
server-side filtering on Exchange without the use of any client-side
software. The performance is pretty good, too, for a non-statistical
filter (about 94% at our site.) iHateSpam Server uses white listing (it
picks up your Exchange address book and your contacts folder) plus a
scoring system for different tokens (probably derived statistically from
some pubic corpus). They up date their "spam definitions" frequently.

Anyway, the reason we picked iHateSpam Server over SpamBayes (and every
other commercial filter, Bayesian or otherwise) is the user interface:
three mail folders are created in each users inbox, one for spam, one
for a white list, and one for a blacklist. With this setup and
everything happening server-side, you can manage the filter from any
mail client (Outlook, OWA, or some random IMAP client). 

Of course, I use SpamBayes on my own mailbox, and performance is better.
But you cannot beat the manageability of an integrated server-side
product like iHateSpam SE. For a long time I've wondered how I could
adapt the SpamBayes code base to work with Exchange server-side, but it
seems like too big a project for me as an individual or even with the
other developers at my company. I wouldn't know where to start.

If any bigger and better group of coders out there would like to tackle
such a SB-Exchange Server project, please let me know and I'll help as
much as possible, with my time and maybe even organizational dollars.


More information about the Spambayes mailing list