Anti-Spam

[NOTE: If you are using POP rather than IMAP or webmail, set your email client to leave mail on the server - you need it to stay there for training purposes]

1) Turn on SpamAssassin in cPanel. Each email now gets a SA header with a score, details, etc.

2) Create an empty file in your account root called ".spamassassinboxenable". This will cause mails flagged as spam to go to a folder called "spam". If the folder doesn't exist, it will be created when the next spam comes in.

3) Use webmail to create two new folders in your email account: "learn_ham" and "learn_spam"

4) Create two cron jobs in cPanel as follows:
15 8 * * * nice -n 15 sa-learn --spam /home/[domain.com]/mail/.learn_spam/cur
20 8 * * * nice -n 15 sa-learn --ham /home/[domain.com]/mail/.learn_ham/cur

Now every day, SA will scan anything in the learning folders and teach itself about your email preferences. Please change [domain.com] to your domain, and check the folder locations.

What must YOU do? Via webmail or an IMAP email client, go through both your "inbox" and your "spam" folder periodically. Move/copy all good mail to "learn_ham" and all bad mail to "learn_spam", regardless of whether it was identified correctly. DON'T just try to train on mistakes.

Until you have trained using a few hundred messages, the Bayes filter module will NOT be active (it would make too many mistakes without sufficient training). So don't be discouraged at first. When the Bayes filter kicks in, the accuracy should suddenly get better, and as you train more, it will REALLY improve.

Once working, you should still train as much as possible since spam tactics/styles change over time. But you can be more relaxed about it, and it will get easier. SA will actually "train itself" (called autolearning) when it sees an email that is very spammy or hammy. But don't rely on that to keep the filter accurate.

Author: Support Desk

Rate this article

How Helpful was this article to you?

• Currently 0.00/5
• 1
• 2
• 3
• 4
• 5

Comments

There are no comments

Add a comment

Name:

E-mail:

Comment

Enter the code: