[LINK] Spam filter eliminates 99% of junk mail

Ivan Trundle ivan at itrundle.com
Fri Nov 16 16:35:12 AEDT 2007


On 16/11/2007, at 11:04 AM, Tom Worthington wrote:

> I did a quick check to see if my SpamAssassin Spam filter was  
> working. The result was that the filter is catching 99% of the Spam,  
> which makes up 96% of the email I receive. No messages were  
> incorrectly classified as Spam.

Interesting data, Tom - it confirms that my system is not that  
unusual, nor special.

> <snip>
> I then looked through the remaining 32 messages not rated as Spam by  
> the system. Of these, seven I considered to be Spam.
>
> The Spam filter appears to be working very well. I could make it  
> more sensitive, but seven junk messages a day is acceptable.

I use DSPAM (from Jonathan Zdziarski, though he has sold the rights  
recently, I gather), and it tells me that it has been 99.759% accurate  
in the past 12 months.

91.393% of my mail is spam: an average day lately indicates that I  
receive about 950 messages as spam. No good messages have been  
classified as spam for a very long time (12 months or more), and spam  
that makes it through to my pop server averages about 2 or 3 per week  
at most.

The settings that I use are the average settings, and I have not had  
any need to make them more or less sensitive, though I do enable  
'noise reduction', and have it automatically whitelist frequent  
correspondents. I have an additional spam filter on my mail  
application, and it successfully identifies the strays. In all, the  
combination of DSPAM and Apple Mail app's junk mail filter capture  
100% of my spam.

I have noted a large increase in spam in the past month - my daily  
average of spam caught by DSPAM usually hovers around the 600-700  
messages per day. And although I've attempted in the past to  
reconfigure my mail server to authenticate mail prior to arrival,  
there are too many government departmental mail systems that I deal  
with which then get blocked. DSPAM adds around 0.416 seconds  
processing time to each message, and offers many other stats on  
everything from inoculations, auto-whitelistings, spam/non-spam, spam  
misses and false positives.

Colleagues who have used or continue to use DSPAM have varying results  
(I manage their domains to a grater degree) - in particular, those who  
like to receive a lot of automated messages heavily laden with non- 
plain text seem to suffer the most (their accuracy fluctuates between  
90% to 98%).

Of course, 99.9% accuracy is all well and good if the spam levels are  
low, since the final messages that get through are the ones that get  
in the way. I'm happy that I rarely see spam these days, and the  
diligence in training DSPAM has thoroughly paid off.

I don't bother with virus protection, as I have no need of it at the  
moment.

iT



More information about the Link mailing list