Morning Campers!

So, the spam situation was getting a little extreme. With 27669 between the start of this month and up to midnight this morning, I'm averaging 1.75 spams per minute. Real e-mail tends to get a little lost amongst that. While my own filters could filter about 50% of the spam out a few months ago, I was lucky if it was getting 5% of it as definite spam, and probable real mail was regularly hitting 1000 emails a day - needless to say, most of them were spam.

A couple of days ago, I set up my mailserver as a mail relay, but I found out today that most of my e-mails I had sent weren't getting through. Eventually tracked it down to an IP blacklist, who had blacklisted the whole of pipex, which means I'll have to fiddle with that a bit more. However, while wading through my spam looking for the returned mails, I found an e-mail from a friend I hadn't heard from for ages. Pure fluke, if I hadn't been in there, I wouldn't have noticed and would have just junked it.

So I decided that enough was enough. I had set up my mail server with a view to dealing with spam at a later time. That time had now come; it was time to take some action. For those of you who don't know much about computers, I drew a picture of what that action may look like:

For those of you who do know about computers, I have installed SpamAssassin. Haven't trained the Bayesian filter yet, but I've been running it out of the box for about an hour with an impressive 93% success rate! 80 spams junked, 6 let through. Now, 144 spams a day I can handle, and with some decent Bayesian filtering, hopefully I can make it even better :)

The only problem with Bayesian training is that they suggest training it with 1000 spams and 1000 non-spams. The 1000 spams obviously aren't a problem, but I don't have many friends, how am I supposed to get anywhere near 1000 for the ham?! I don't even get near it when I add my hate mail! Damn the server crash of '03, I'd kept every email going back to when I came on the internet. That would have got me 10th of the way there...

Still, I'm feeling positive, and hope that the battle over my inbox has taken a turn in the right direction.

In other news, today has been my first full day at home with nothing to do since, er, last summer, how nice. I spent it speccing up a Doom 3 machine... Well, this one is 3 years old now, time for an upgrade! Also watched a video about halflife 2 - mmmm, it looks good.

Comments

We found that bayesian filtering was pretty much rendered useless by the bayes poisoning stuff that the spammers have been sending out of late. It's probably more effective with small amount of personal mail though.

I can't recommend enough http://www.rulesemporium.com - grab the rulesDuJour script, and use the big list it has. Those *vastly* improved our accuracy.

Turning on spamassassins DNSBL support is also worthwhile...

Leave a comment