Slashdot Mirror

I'm not so sure what everyone is complaining about. I'm using SpamSieve as a plug in to Mail.app, and it catches just about everything without much in the way of training. Currently, my statistics as of 2006-11-01 say it's 97.1% accurate (with 71% of my total mail volume being spam, but that includes some legitimate marketing mail that I no longer really want, and I'm too lazy to track down the list maintainers), and that number gets higher every day.

On Windows, I'm using either Mozilla Thunderbird (usually), or SpamBayes as a plug-in to Outlook 2003 (when I have to), and I get similar results.

Of course, what we really need to do is rethink the way that the whole email system is designed, just in terms of MTAs that work separately from MDAs, etc. This kind of filtering really needs to take place at what we currently call the MTA level, with a configurable corpus for each user. The filtering should be done before the mail is permanently accepted, so that the impact on storage resources is as minimal as possible. Granted, it still takes a lot of processing power.

Another thing I need to spend some time thinking about is how RFC822 messages are structured in general. I'm just pulling this out of my ass right now, but the fact is that message envelopes are much to easy to spoof. Why have a separate message envelope to route the mail when the addressing information is already supposed to be contained in the headers? With the way spam is going, the message needs to be processed in its entirety in any case, so perhaps the envelope has outlived its usefulness?

SpamSieve has been missing a lot of those lately for me. Getting better but still missing more than I'm used to. However, a new version was released this week which has as its first feature better detection of these. Woo hoo!

SpamSieve has been missing a lot of those lately for me. Getting better but still missing more than I'm used to. However, a new version was released this week which has as its first feature better detection of these. Woo hoo!

Lots:

iGet (geat file transfer tool) is alread shipping as a Universal Binary.

BBEdit (killer text editor) is alread shipping as a Universal Binary.

SpamSieve (best Bayesian antispam) is alread shipping (its plugin) as a Universal Binary.

OmniWeb (world's best browser) is in testing (likely to be ready immediately upon Intel Macs shipping).

And many others, etc. etc. etc. Only the huge, slow-cycle behemoths will really be lagging. But the real question is, when will FinderPop go Intel-native? That will dictate my own Intel migration timeframe...

If you're using OS X Mail there are alternatives to setting up a ton of manual rules in the prefs.

Junkmatcher is a good spam matching plugin and SpamSieve is a great Bayesian filter that does any amazingly good job -- even with otherwise legitimately looking mail.

I'd like to second SpamSieve. If more than one piece of spam gets through in a day (where each day I receive > 500 pieces of email), I am truly surprised. My stats for June are:

• 1007 Good Messages
• 13729 Spam Messages (93%)
• 1 False Positives
• 24 False Negatives (96%)
• 99.8% Correct

Works for me. Oh, the false positive was a list that I just signed up for. They sent a confirmation mail, I checked to see if it was caught (it was), and marked it as "good". Piece of cake.

You can have Bayesian filtering in Mail, with SpamSieve from Michael Tsai.

You might also be interested in reading Joe Kissell's just-released ebook Take Control of Spam with Apple Mail, which explains the common accuracy problems with Mail's Junk filter and how to optimize it for better results. Joe also recommends SpamSieve as an alternative to Mail's Junk filter in those instances where Mail proves inadequate.

cheers... -Adam

You matter how good your filter is, it seemes as if it always makes some mistakes.

It might be a better idea to rate spam by what the odds are that it is spam. Then you rate all you email but how sure you are that it is spam. Start reading your mail till you get to the spam, after that you can be pretty sure everything that comes after it is also spam. You get to define the final boundry, the ratings just help you find it.

Spam Seive, and other filters come close but still still try eliminate any grey area so you will always have false positives, and spam that gets through.