SpamSieve 2.4
A updated version (2.4) was released today, featuring the following major improvements:
- Made various changes to the Bayesian engine to improve accuracy.
- Added some heuristics for detecting phishes.
- Apple Mail messages can now be filed into different mailboxes based on how spammy they are (requires 10.3 or 10.4). (Read more about this new feature)
- The Apple Mail plug-in is now a Universal Binary, so SpamSieve can be used on Intel-based Macs without running Mail in Rosetta.
You can read the full Changelog on the SpamSieve web site.
SpamSieve is shareware (USD 25) and is available from the developer’s web site.
Similar Posts:
- Mail.app’s Junk Filter is not like the others
- SpamSieve 2.4.1
- How Bayesian spam filtering works
- JunkMatcher: free extra spam protection
- SpamSieve gets Thunderbird support
Tags: Apple Mail, Junk, spam, SpamSieve
December 21st, 2005 at 10:14 am
[...] JunkMatcher or SpamSieve. It’s a draw for me which of these offers the better additional protection from spam. [...]
January 26th, 2006 at 8:27 am
[...] SpamSieve (additional Bayesian spam-protection) joins the list of apps recompiled as universal binaries. [...]
January 27th, 2006 at 7:34 am
Another application that has become pretty popular (yet it is not mentioned anywhere on Hawk Wings!) is POPFile, which is essentially just a Perl script, so it is very portable. (personally, I haven’t used it because I haven’t had a need, but I hear all good things)
http://popfile.sourceforge.net/
This application does much more than Bayesian filtering for spam. It does Bayesian filtering for complete automatic e-mail classification.
Think of it this way — why train your filter only to find spam? Why not train it to classify all of your mail?
Imagine that you are a professor for a class. You have lots of students e-mailing you questions; however, the way they identify themselves and the classes that they are in vary widely from message to message so you cannot setup a standard filter to filter them into a course box. Well, let POPFile use its Bayesian filter to determine the probability that a message should be sorted into a course box and, if the probability is high enough, actually do the sorting.
What’s more fun is that POPFile simply tags messages. It adds a header that you can use a filter to catch later.
And don’t be deceived by the POP in POPFile. From what I read, it can be used as an IMAP frontend as well.
It’s free too.
January 27th, 2006 at 8:11 am
Hi Ted.
Don’t take this wrong way. I am in awe of your posts – cronjobs, procmail scripts and all that hardcore goodness ;-)
But here I am wondering: Mail.app’s rules are already sorting my mail and the Junk Filter catches 99% of my spam.
What does POPFile add to that? Am I missing something?
January 27th, 2006 at 8:24 am
Couldn’t you make the same argument about SpamSieve? If Apple Mail is good enough for you, why show interest in SpamSieve?
To me, what’s neat about POPFile is the fact that it uses the Bayesian machinery to setup novel filters that you simply cannot setup with Apple Mail (or Thunderbird or any other) filters. POPfile takes Paul Graham’s original idea for spam filtering and extends it to the rest of your mailbox.
Spam, for example, cannot be captured by a simple filter. You need a filter that can grow and adapt and change and look for things that aren’t captured by most simple filter machinery. Thus Bayesian (and related) spam filters build nebulous descriptions messages captured by the synergy of dynamically updated prior probabilities and special heuristics/fuzzy-logic.
Aren’t there other types of messages in your life that cannot be described by a simple filter? How do you describe a “work” message? What if you could classify it before even reading it? Would that improve your productivity?
So that’s what POPFile adds. Rather than setting up a simple filter based on subject and sender, you setup buckets and don’t give any other information. Gradually as you start classifying messages as belonging to certain buckets (and from what I understand, buckets can be thought of as tags, so there’s no reason messages have to go into only one bucket) it builds your filter for you.
That being said, I don’t have much desire to use POPFile because I don’t have the need for advanced e-mail classification. However, if I did (and some people do) I think that POPFile provides a novel solution.
January 27th, 2006 at 8:26 am
What I do think would be nice is if someone would extend this idea to blog publishing software and tags. (maybe they’ve already done this?)
Wouldn’t it be great if blog software was able to pick relevant tags FOR YOU as you type your post? This is similar to what del.icio.us does when you tag a new page and it suggests new tags. Once you get to the end of your post, the software clicks the tags it thinks you want, then you correct its choices, and it “learns” from you. del.icio.us doesn’t do a GREAT job picking tags for me when I tag pages, but it’s trying to learn what the “average” person is like. A personal web blogging software should have the ability to hone in on my tag preferences pretty well.
January 27th, 2006 at 8:44 am
Ted wrote:
I think you might have me there. I surrender. :-)
I’ll blog it up.
This is particularly interesting:
June 20th, 2006 at 2:26 am
In regards to the POPFile comments. I’ve used POPFile extensively for several years now on the mailserver level and client level and it’s results are utterly amazing. On a server with 5 domains and a multitude of users I have been able to easily matain a spam detection rate of no lower than 99.7% with negligable false positives/negatives. The only baysian product I have found (and I have tested many) that is comparable to POPFile is K9. K9 is actually a windows maternal twin of POPFile, using the same baysian setup but is coded for windows specifically and is super for windows end users. It doesn’t do SMTP though, which makes it unusable on a mailserver. It produces the exact same effectivness results as POPFile. I have found no other baysian based spam detection tool that can produce results realized in POPFile/K9. I imagine most will do fine for someone with minimal accounts and rely on whitelisting/blacklisting to take the bulk of spam processing off their baysian filtering but I defy any develop to claim that their baysian filtering setup can rival POPFile’s or K9′s accuracy.
The one drawback to POPFile is that it’s not multi-user, multi-database and all correction have to be made to one database. On the server level that spells trouble. The system administrator must manually make corrections for users. It’s either that or the admin must allow the users to make database corrections to the central database and any admin that allows users access to that needs to be taken out and put up against the wall.
POPFile does have a mutil-user version in development, however, it’s been in developement for a long. long while and it’s more than a year beyond it’s projected release date. I often wonder if the multi-user version will ever become a reality.
January 9th, 2007 at 3:09 pm
I’ve been using SpamSieve for a couple of months on an e-mail address that I’ve had for about twelve years. It gathers a lot of spam. My stats (on SpamSieve) show that I’m averaging about 140 spam messages every day and the program is catching 99.8% of them. Before SpamSieve I was having to deal with spam every single day (in spite of the Mail’s built-in spam filter), while now I hardly ever see a single piece of spam directly. It’s great. I haven’t seen anything else come anywhere near doing that good.