Quick update on SpamBayes: I FREAKIN’ LOVE IT!. So far I’ve tried a grand total of two client-side spam blockers – Norton Antispam, which came installed with a six-month trial on my new Dell at work, and SpamBayes, which is open source and free for all.
I won’t mince words here – Norton sucked. In fact, I’ve noticed more and more that their offerings have been going very downhill as of late. Norton Antivirus takes up entirely too much resources, especially when compared with TrendMicro’s offerings. Systemworks doesn’t seem to fix as money problems as it used to. And the Antispam program, which I had to physically enable even though I’d installed it, didn’t do a thing to my number of spam.
SpamBayes is the day to Norton Antispam’s night. I’m now using both the Outlook Plugin version and the POP/SMTP proxy through Outlook Express. Both are laughably easy to set up. The plugin is far easier to train – just put all your existing spam in one folder and all your good mail in another, tell the program where each is and let it go. The proxy required me to set up a rule in Express to forward everything from each folder to the local fake email addresses I set up (to train the proxy, you send bad spam to “spambayes_spam@localhost” and good mail to “spambayes_ham@localhost” and it grabs them before the SMTP server sees them).
After the plugin version is trained, it dumps anything identified as spam in your “Junk Mail” folder and anything it’s unsure about in the “Junk Suspects” folder. In the last seven days, I have received a whopping 474 pieces of Spam mail. Only about 10 of them slipped through into my good email and maybe 50 – 75 have ended up in the Suspects folder. Of my good mail, only about 20-30 understandably questionable, but still good by my standards, emails ended up in the Suspects folder and absolutely none ended up in the Junk folder.
The proxy works a little differently. Instead of assigning emails to folders (which isn’t possible in POP from the server side) it signifies it’s impression of each incoming mail by adding “spam”, “unsure” or, optionally, “ham” someplace in the message (I set it up to place the word at the beginning of each subject). It’s then up to my mail client to sort through them. The results on the proxy have been roughly the same as with the plugin so far. I’m truly impressed.
What most interesting to me is that, now that my email is organized this way, I actually read some of the spam I get. I have absolutely, positively no intention of ever purchasing anything from it, but I find my attitude toward spam changing. I used to angrily delete or simply ignore all the spam sitting in my inbox. I was infuriated that these people were wasting my time and cluttering my email with useless crap that had absolutely no relevance to my life. Now that they’re regimented to different folders automatically, however, that hatred is gone, and it’s becoming sort of entertaining to read their pitches.
The proxy, while easy to install, I’d imagine would be a bit tricky for the uninitiated to configure. The plugin, on the other hand, is pretty user friendly. It installs a button in Outlook labeled “Delete as Spam” and another when you’re in the spam folder labeled “Recover from Spam”. I think it’s pretty easy how you’re meant to continue training it. And, trust me, both versions absolutely improve over time with training.
The downside of this system is that it sits on the client. Essentially, you’ve already downloaded all of the bandwidth-sucking spam onto your mail server and then onto your client computer by the time SpamBayes gets to it. So, while it does clear the clutter that you see, it doesn’t speed up your email connection. In fact, it actually slows it just a bit as the program checks every message and assigns it a score. Of course, you can install the proxy on your mail server, but then you lose the coolness of the plugin’s functionality.
At any rate, if you use POP3 to get your email or use Microsoft Outlook, SpamBayes will save your life from the clutches of spam. It’s much better than the SpamAssassin/Earthlink tactic of acting as a gateway protector, sending automated message to spammy-looking emailers requesting some sort of user intervention. If I already sent you the email and your server rejected it, I’m not going to be too thrilled about having to send a second message or send a web page to prove I am who I am. Whitelists and blacklists are too easy to defeat and are a maintenance nightmare.
The Bayesian method, on the other hand, leaves it to you to determine what is good and bad and learns from your decisions. It’s tailored specifically to you, which makes it that more effective. And while the setup and initial training can be a bit time consuming, the maintenance is a breeze. Just be sure to always tell the filter when a bad email slipped through and it’ll just keep improving itself. How can you complain about software that actually makes itself more effective at its job?