Changing the World With Metrics

[bcholmes]

Since last year, I've been pestering the e-mail guy at work to install Spam-filtering software, since I get inundated with the stuff.

It's pretty bad. My bcholmes.org domain kicks out the most obvious spam, and I probably get about 5 to 10 pieces of spam a day. But work is just ridiculous. It doesn't help that we use the worst of all possible e-mail clients, so there's only so much I can do about, say, turning off images and not running scripts in e-mail.

Some time in May, I started tracking how much spam I get. It was over 120 pieces of spam a day. I kept tracking this in a graph, and would periodically print out the graph and drop in on my boss' desk.

I had the active e-mail address that received the most spam in our office (I figure because I once posted some questions to a Java newsgroup -- my e-mail was harvested from there). There were four inactive addresses that received more spam than me, but I was top of the list for active e-mail addresses.

Some time in August, we installed spam-filtering software, and now I get around one spam a day.

Comments

[futabachan.livejournal.com]

sue@sue.net gets hundreds per day -- I've been on Usenet since forever, and I'm trivially vulnerable to dictionary spams, too. I'm changing email providers soon, and will have filters up the wazoo....

[epi-lj.livejournal.com]

Cute graph. :)

If I disable our filtering entirely, I get about 1500 spam messages a day to my work address. I use two layers of filtering: A server-side filter which covers the whole office but is not set to be *that* tight, and then for people who get really hammered, an Outlook plugin that implements a very, very nice Bayesian filter. My spam rate has dropped from 1500 a day to about ten or twenty, with no appreciable level of false positives at all.

Incidentally, what filtering software are you using? (We're using postfix / spamassassin / razor on the server and SpamBayes on the client.)

I read an article somewhere about disabling image loading in Outlook, which is what I'm assuming the mystery client was, but I'm not finding it offhand right now.

[bcholmes.livejournal.com]

We're using Brightmail Anti-Spam, and yup, the e-mail client is Outlook.

[nolly.livejournal.com]

I've been very happy with SpamBayes at work, too.

[bcholmes.livejournal.com]

So this was the first time I'd heard of Bayesian analysis as a method of Spam filtering. What I found interesting was that when I Googled "Bayesian", Google gave me an ad to work for Google.