Adrian Sutton describes how he has set up SpamAssassin and Mail’s inbuilt Junk Filter to work together for extra spam-catching cleverness.

Having set up both SpamAssassin and Mail to dump their spam in Mail.app’s Junk folder, he runs SpamAssassin’s sa-learn script in a cron job so that it learns from what Mail’s Junk filter catches as well.

Running sa-learn –ham on Mail’s Inbox after returns false positives with Mail’s “Not Junk” button, helps SpamAssassin learn what’s not spam.

As a result, he has “a partially self-training system with a user-friendly interface”.mail.app, apple mail, spam, junk, tips, junk filter, self-training

Spammers have discovered a new trick.

According to a report in USA Today, image-based spam is experiencing a huge growth spurt. Late last year it accounted for 1% of total spam messages; now it has suddenly risen to 21%.

In the newspaper article, journalist Jon Swartz describes the problem:

The newest spam uses technology that varies the content of individual messages — through colors, backgrounds, picture sizes or font types — so they appear to be distinct to spam filters… As a result, the messages are like snowflakes: No two are alike… The surge in new spam has largely eluded software filters and eaten up space on e-mail systems because each message is more than seven times larger than regular spam…

I’ve certainly noticed a recent surge in the amount of spam getting past Apple Mail’s excellent Junk Filter.

Much of it is image-based, although I am favoured more with various aids for my sexual potency than with the stock scams that are usually associated with image spam.

Mail’s Junk filter, based on Latent Semantic Analysis, only gets smarter as it goes along. I hope it is not a slow learner. Although, since the filter is text-based, this new form of spam may have outflanked it.

[Thanks, Bob]mail.app, apple mail, spam, junk filter, images, email, Internet, stock scams

Feast your eyes on this little beauty:

This is Bayes’ formula. It is the maths behind the Bayesian filtering used by spam-catching apps like SpamSieve, SpamFire and JunkMatcher.

Process Software offers a lucid and racy explanation of how Bayes’ forumula works to keep spam out of your Inbox. You will be surprised how interesting it is. I was.

Apple Mail’s built-in Junk Filter uses a different process called “Latent Semantic Analysis”.

“Mail.app’s Junk Filter is not like the others” contains a number of links to explanations of LSA and how it differs from Bayesian filtering.spam, lsa, latent semantic analysis, bayesian filtering, junk filter, mail.app, apple mail

Steve Dispensa tells a funny story at Kernel Mustard:

My Apple mail client has a built-in spam filter. I’m not sure how it works, but it does have to be trained. Anyway, that’s all I know about it.

It turns out that of all of the copious communications I receive from Microsoft, the ONLY thing that the spam filter catches – EVERY TIME – is the Microsoft Mactopia (Mac Business Unit) newsletter. I mark it as “not junk” every time. It refuses to learn.

(For what it’s worth, Mail.app is largely a piece of crap, with tons of un-Apple-like bugs. Don’t get me started.)

False positives are annoying. But my Mactopia newsletter turned up without problem in my Inbox this morning, so it doesn’t have to be this way.

Three ways to reduce false positives:

1. Keep your Address Book up to date. You can set the Junk Filter to let through messages from people whose email addresses are in your Address Book (Preferences > Junk Mail > Sender of mail is on my Address Book), but it only works well if all the addresses you value are in it. (Could this be Steve’s problem?)
2. Make sure that you always correct the Junk Filter if it incorrectly marks a message as Junk Mail by clicking the “Not Junk” in the Toolbar or hitting Command-Shift-J. This ongoing learning stops Junk Filter making the same mistake more than once.
3. If nothing else works, create a rule for email that Junk Filter keeps wrongly marking as spam. Add the “Stop evaluating rules” action to the rule and Junk Filter will ignore the email.

These excellent tips are taken (shamelessly) from Joe Kissell’s “Take Control of Spam with Apple Mail” ebook. Buy it .

It makes an first-rate companion to his equally-useful Take Control of Apple Mail in Tiger ebook. Or save money and buy them both .

AOL has released the ten most common subject lines in spam emails for 2005.

“Donald Trump” and “penis patch” top the list.

Normally I don’t get to see these due to the efficiency of Mail.app‘s Junk Filter, so the list is news to me (as I hope it is to you):

1. Donald Trump Wants You – Please Respond;
2. Double Standards New Product – Penis Patch;
3. Body Wrap: Lose 6-20 inches in one hour;
4. Get an Apple iPod Nano, PS3 or Xbox 360 for Free;
5. It?¢‚Ǩ‚Ñ¢s Lisa, I must have sent you to the wrong site;
6. Breaking Stock News** Small Cap Issue Poised to Triple;
7. Thank you for your business. Shipment notification;
8. Your Mortgage Application is Ready; Thank you:
9. Your $199 Rolex Special Included;
10. Online Prescriptions Made Easy.

Commenting on the overall nature of spam in 2005, AOL postmaster Charles Stiles suggests:

While the volume of spam reaching AOL email inboxes has remained at low levels compared to its height in late 2003, the spam that’s out there is more insidious, crafty, devious, and dangerous than ever.

[Quote via InformationWeek, list via the 168 Media Group. This is an example of attribution, citing the sources for a post. So honourable. So quaint.]

Mail.app‘s Junk filter is not a Bayesian one like the spam catching technology in SpamSieve and many other spam-fighting apps. Instead it uses a technique called Latent Semantic Analysis (LSA).

I’ve blogged two excellent descriptions of the way this filter works in Apple Mail before, but today came across a third explanation with the imposing title, “Bayesian Nets, Latent Semantics, Despamming and other speculations”.

The post describes the differences between Bayesian and LSA and a whole lot more. It’s over two years old, but a good read and packed full of links to other resources on spam.

[via Paul's Time Sink.]spam, bayesian filtering, Latent Semantic Analysis, LSA, mail.app, apple mail, junk filter

It’s a quiet time with not much to blog. So I entertained myself by spring cleaning my Previous Recipients list.

Apple’s tech note on how to keep your Previous Recipients list up-to-date is one of my Top Five Apple Tech Notes for Mail Users.

You can find the list by selecting “Previous Recipients” from Mail’s Window menu.

It’s a good idea to go through the list from time to time because:

1. It stores incorrectly typed addresses and keeps suggesting them via Mail.app‘s auto-complete feature. That can be a pain.
2. It stores every address for every email you have ever sent, even addresses that people haven’t used for years. Do yourself a favour: Clean out the old ones.
3. Apple Mail‘s Junk Filter will not mark an email as junk if it comes from an address in your Previous Recipients list. Along with the Address Book, it functions as a de facto “white list“. Keeping it up to date helps Mail to find junk better
4. It’s quite fun to go through the list and wonder who all these people are that you have emailed at least once. It’s certainly more fun than spring-cleaning the house.