Just read about some interesting new spam filters. Slashdot states [1]
that the filters have reached accuracy levels between 99.983% and
99.984% (1 misclassification in 6250 messages).

Perhaps you should check out CRM114 [ http://crm114.sourceforge.net/ ]
and DSPAM [ http://www.nuclearelephant.com/projects/dspam/ ].

[1] http://yro.slashdot.org/yro/04/02/24/0025219.shtml

-JohnF

John Marrett wrote:
>>- anti-bayesian spam : many message whose plain-text part is
>>"cucumber democracy outrage xcfqsd royal peanut weed ...."
>>that drove my bogofilter nuts. Shall I remove the '-u' option
>>that tells it to register messages that have just gone through ?
>>Maybe it should only read the HTML part when there is one ?
>
>
>
> Most of the spam that I receive these days has anti-bayesian word lists
> included. I am using SpamBayes and it is blocking 95%+ of the spam that I
> receive. The other 5% is considered suspect. I haven't seen a spam message
> in my inbox in months.
>
>
> I use the outlook plugin (work requires outlook) so I train my filter with
> every messages. Are you training both ham and spam messages? If you aren't
> training enough ham messages it could be affecting the effectiveness of
> your filters.
>
>
>
>>- bounces. It seems I was in many address books, as I'm
>>receving hundreds of bounces for virus that spoofed my addres.
>>Provided there is only one coputer I send mail from, do you
>>think I should try to remeber all the message-ids I sent and
>>make a filter based on that ?
>>
>
>
>
> If you are willing to discard all virus warnings you could create filters
> based on the Antivirus responses. If you block the standard responses for
> the top 5 products you should block most if not all virus responses.
>
> I think that remembering all of the message-ids could get out of hand vey
> quickly.
>
>
> -JohnF
>
>