Spam filters

[time]

Over the last couple of weeks, I've been trialling a few different email clients, particularly with regard to spam filtering. I don't get much spam compared to many people, but it's a growing problem.

I used Spamhilator with my regular email software for comparison. Here's my rankings so far:

1. Thunderbird 0.8. The filter was slow to learn, but increasingly accurate. I'm impressed.

2. Foxmail 5.0. Just a whisker behind TB, yet this is without Bayesian filtering, which doesn't kick in until you've accumulated 1000 messages each of ham and spam! The automatic rules-based filters worked well right from the start and just get better and better. I'd assume that once the Bayesian filtering kicked in, this product would be awesome.

3. Spamhilator. A fair way behind the first two with both false positives and negatives, but satisfactory.

4. Barca (Poco Mail). A sophisticated Outlook alternative with scheduling etc - pity it's got version 1 syndrome (bugs, missing options and no doco). The spam filtering started fairly well but drooped as time went on, despite my efforts to tweak it. I hope they can improve it; it's a nice looking product.

5. Opera (M2). Not too bad to start with, but I couldn't work out how you're supposed to change spam back to ham when wrongly identified, so I lost interest. I included it because I know a couple of people who use it, plus it uses standard Mbox storage.

6. iScribe. Unfortunately, the version I tried had a bug that stopped it learning to identify spam at all. The latest version just crashes when I try to do anything with spam messages, and I couldn't find any intermediate versions of the software. It's quite possible that my data is corrupt. A shame, because Scribe is the only product that has the wit to keep suspected spam separate from confirmed spam (apart from Spamhilator). It's really quite tedious peering at a long list of spam checking for what's new.

Only Spamhilator allows you to properly convert spam to ham when it's been misclassified. The others let you flag it easily enough, but fail to move it from the junk folder back to your inbox. Very irritating.

What's other people's experience?

 

[Mercutio]

The reason you just gave is the reason I install and support Spamihilator on client machines. It works with a minimum of fuss, is easily tunable to your needs and is simple enough that non-technical users can handle it. It's a minor annoyance to have to check the spambox every so often for useful mail, but as time said, at least when you flag something as "good", it goes right to your inbox.

I've got thunderbird working well on my PCs. It's a happy thing for me.

Anyone used Outlook 2003's filters? I'm just curious if it's a real feature or just something they put in their marketing literature.

(I won't even make fun of you for using outlook... at least, not in this thread).

 

[Fushigi]

I've been using Spamhilator + Outlook Pro 2000 at home for some months now. Overall I like the combo, but Spamhilator needs some additional flexibility / rules capability to make it a hands-down recommendation.

I have 2 main accounts at my domain + an "anyone at" account for everything else. Outlook checks them all and uses rules to sort them to different folders. Makes it easy for on-the-fly accounts to use for rebates & other misc. uses.

One nice thing is that for suspected spam that made it through, I can go into Spamhilator and have it learn from the message. Also nice is I can view the message as raw text there vs. opening it in Outlook - eliminates macro virus issues, tagged graphics, etc.

What I don't like is the subject field doesn't seem to get passed through the stop-word list. This is bad as people/bots are always creating dummy accounts against my domain and the spam they generate never falls off. My "anyone at" account gets 80-250 messages a day. I could significantly reduce the spam checking overhead if the stop-list would check the subject lines. I could then include variations on male potency drugs and the names on the accounts. To be fair, I haven't suggested this as an improvement to the product. I think I'll send off an email later today...

 

[sechs]

I've found the junk filters on Eudora to be surprisingly good, and, more importantly, easy to use.

 

[Will Rickards]

Thunderbird is catching most of my junk mail. (7500 since may)
The only ones that seem to get through are the really short ones that look like someone is sending me a link to a site I requested.

 

[Fushigi]

I just now went and downloaded a few more filters for Spamihilator. Maybe that'll ease the burden some. Just noticed the Oulook add-in to flag senders as friend or foe from within Outlook - nice!

To firm up what I mentioned above, Spamihilator's stats show an average of 192 spams/day.

 

[Handruin]

Anyone used Outlook 2003's filters? I'm just curious if it's a real feature or just something they put in their marketing literature.

(I won't even make fun of you for using outlook... at least, not in this thread).

I'm using outlook 2003 and their specific junk e-mail filter does not do a good job. I've tried to make it learn several hundred spam messages, and it only works about 30% of the time. After reading the online help (just now), I've discovered their is an option to increase the level of spam filtering. (the wonders of online help)

I would rather use thunderbird if it worked better with all my IMAP accounts. I've complained in another thread of a performance issue.

About the [outlook 2003] Junk E-mail Filter
The new Junk E-mail Filter replaces the rules used in previous versions of Microsoft Outlook to filter e-mail messages. The Junk E-mail Filter feature is on by default, and the protection level is set to Low, which is designed to catch the most obvious junk e-mail messages. Any message that is caught by the Junk E-mail Filter is moved to a special Junk E-mail folder, where you can retrieve or review it at a later time. You can make the filter more aggressive, which may mistakenly catch legitimate messages, or you can even set Microsoft Office Outlook 2003 to permanently delete junk e-mail messages.

There are two parts to the Junk E-mail Filter: the Junk E-mail Lists and the state-of-the-art technology developed by Microsoft Research that is used to evaluate whether an unread message should be treated as a junk e-mail message based on several factors, such as the time it was sent and the content of the message. The filter does not single out any particular sender or type of e-mail message. The filter is based on the content of the message in general and uses advanced analysis of the message structure to determine the probability that it is a junk e-mail message.

 

[blakerwry]

My work subscribes to postini... absolutely the best server side filtering I've ever seen...

works so well that I do not need any spam filtering software on my PCs.

Anybody tried mailwasher? It was recomended to me, but I've never used it.

 

[mangyDOG]

If you use MS Outlook, I have found SpamBayes to be almost perfect, and if you have kept a pile of spam in your deleted items you can train it instantly. I haven't had a single legitimate e-mail deleted since I started using it and only get one or two "maybe spam" messages per week, all the rest are nuked.

If you don't use Outlook you can still use spambayes with a webbrowser interface but it is not quite as easy to use, but just as effective. The webbrowser interface can also be used as a spam filter frontend for a mail server program.

Best of all it is completely free (Donation ware only).

Get it here = http://spambayes.sourceforge.net/windows.html

cheers,
mangyDOG

 

[time]

Update:

My iScribe data was indeed corrupt. I installed it from scratch and now it's easily matching Thunderbird. They're both pulling away from Foxmail.

I misunderstood Opera M2 spam functionality because of its 'View' paradigm (you drag received mail into one or more views, but the original stays in the Received list). The spam folder is actually another View, so you just drag missd spam into it. They provide a "Not-Spam" button to remove it from the Junk view because it doesn't make sense to drag it anywhere - it's still in the Received list. This works better than anything else here in dealing with ham incorrectly identified as spam.

Anyway, it detects reasonably well - about the same as Spamhilator I reckon. I haven't been able to fault the client otherwise. Only it and Thunderbird were able to render an inline html doc I just received. It's not for me, but I don't think I could come up with reasons for someone to change to something else.

 

[time]

Anybody tried mailwasher? It was recomended to me, but I've never used it.

It only looks at headers. So if allowed, it will happily delete ham off your server that it thinks is spam! This is quite unlike proper server-side software in that you have no chance for review.

If you choose to review spam, it downloads the message bodies anyway, thereby cancelling any bandwidth advantages. I couldn't see a huge advantage over using a suitably equipped mail client to browse the headers directly on the server.

It relies heavily on automatically adding addresses to a black list, which of course is largely useless for most spam.

What really raises my ire is that they actively encourage you to bounce the spam! Makes you wonder whose side they're on ...

I think people like Mailwasher because they feel empowered. After all, it's quite a leap from being treated like a mushroom by Outlook Express.

 

[ddrueding]

I use Outlook 2003, I noticed the "Junk Mail" folder...nothing's gone into it yet. I run SpamAssasin on my Exchange server and about 1 spam per week gets through to my inbox. I have in the last 3 months had 2 false positives. This includes all my addresses and a catch-all box, it deletes about 100 messages a day. With all it's bells and whistles active and a threshold score of 7 it's a dream.

 

[Fushigi]

There's an update to Spamihilator that adds new filtering capabilities. Use the "check for updates" function to d/l & apply or go straight to http://www.spamihilator.com/ .

The new Learning Filter (Bayesian Filter) uses the rules of Thomas Bayes (English mathematician, 18th century) and calculates a certain Spam-Probability for every E-Mail. You can train this filter! So it will know your messages even better than you. Hence the recognition rate will continuously increase.