How to Reduce Spam



As spam recipients go, I'm a class act. I get hundreds every day and on some days, more than 1,000. Yet in my mail box I hardly see any. Here's how I do it.

First, a little history. I used to use a Bayesian spam filter running on my PC. The product I used was an Outlook add-in called JunkOut. Like most Bayesian filters it took a while to train but once trained it worked just fine. The spam detection rate was around 98% and the number of false positives (good mail wrongly classified as spam) was vanishingly small.

But there was a problem. As my spam mail volume grew, the time taken by the spam filter to process my mail was growing to the point of being unacceptable. Some days it was taking 10 minutes or more to process my mail.

I needed a different solution. I tried rule-based spam filters that used less processing than Bayesian filters. I tried setting up my own mail server on a dedicated PC. I tried various commercial spam filtering services and other options as well. None of these gave me what I was looking for. But then I tried Google's GMail and bingo! I found what I had been looking for.

Unlike some other webmail services, Gmail provides spam filtering for free. That's no big deal; Yahoo!, Hotmail and others do that as well. What's different about Gmail is that it also provides free POP3 mail access.

Most of my spam mail is sent to the address That's no surprise; that address appears in every issue of this newsletter and on my website as well.

What I do is to forward all mail from that address to my Gmail account where it is spam filtered automatically.

The GMail spam filter detection rate is good, around 95%, so around 950 of the 1000 spam messages I receive daily never get to my Google Inbox.

I then use POP3 access to download the contents of my Google Inbox to Outlook. The incoming mail is then filtered using the excellent network-based Cloudmark spam filter that is installed on my PC.

Cloudmark's detection rate is around 92%. So of the 50 or so spam emails in my Google Inbox each day, fewer than five make it through to my Outlook Inbox.

Now here's the crunch. Both the Google GMail spam filter and the Cloudmark spam filter have the same characteristic; they virtually never classify my real mail as spam. That means I don't need to regularly check my spam folders to see if they contain genuine correspondence. That's a real plus with large spam folders.

The spam detection rates for GMail and Cloudmark are good, though a long way from the best in their class. But that doesn't matter. By chaining the two systems together I increase my aggregate spam detection rate to 99% plus and that rate IS right up there with the best.

So the end result is that of 1,000 spam emails per day I see fewer than five. At the same time my real mail is virtually never sent to a spam folder. Problem solved.

Forwarding my to GMail is easy for me as I control the email to my website. Other users, however, may not have the facility to redirect mail from their normal mail account to a GMail account. Some mail services provide this feature, others don't; you'll have to check your service to find out.

Even if your account doesn't allow mail forwarding you can do it yourself using a free utility called ERC. This runs on your PC and can be scheduled to automatically log into your mail account and forward the mail to another account. In fact, it can forward mail from up to three different accounts.

A better solution in the long run, though, may be to shift your permanent email address to Gmail.

I use Cloudmark for my secondary spam filter but a good Bayesian filter would perform well in the role provided you have the patience to train it. SpamBayes and K9 are good examples and both are free. The Thunderbird mail client of course has its own built in Bayesian spam filter so there is no need for another.

Whatever product you chose, I do suggest you try this combination of remote and local spam filtering. It could be just what you have been looking for.

Cloudmark Desktop:



Please rate this article: 

Your rating: None
Average: 3.5 (24 votes)