Spam Aversion Technology

How We Protect Your Inbox from Spam

We use a layered defense system to stop junk mail before it ever reaches your inbox. Our system works like a team of specialized security guards who examine every incoming email. The core components of this system are Real-time Blackhole Lists, Amavisd, SpamAssassin, and a special kind of artificial intelligence called Bayesian Filtering.

1. Real-time Blackhole Lists (also called RBLs)

Millions of spam messages are identified and their origins are flagged. When an email server connects to deliver messages to our users, the system checks the sever RBLs to see if the sender is on the list. If it is, the message is declined.

2. Amavisd: The Traffic Cop and Coordinator 

Think of Amavisd as the chief traffic cop or delivery supervisor for all your emails.

• Its Job: When an email first arrives, Amavisd steps in. Its primary function is not to check for spam itself, but to coordinate the process. It ensures every email is properly handed off to the virus scanners and the spam checker (SpamAssassin).
• The Flow: Amavisd takes an email, sends it to the other tools for inspection, waits for their results, and then decides where to route the email next (to your inbox, the spam folder, or blocked entirely). It makes sure every security check is performed efficiently and in the correct order.

3. SpamAssassin: The Rule-Book and Scoring System 

SpamAssassin is the lead inspector that runs hundreds of tests on every single email and assigns it a spam "score."

• Its Job: It checks an email against a massive, constantly updated set of rules. It looks for everything from known scam phrases and suspicious links to technical flaws in the email's structure (like a fake sender address).
• The Score: Each failed test adds points to the email's score.

• High Score: If the score is too high (e.g., it hits our "kill" limit), SpamAssassin tells Amavisd to reject or quarantine the message.
• Medium Score: If the score is moderate, the email might be marked with ***SPAM*** in the subject line but still delivered to your inbox or Junk folder.

• Examples of Tests:

• Rule Check: "Does the email contain the words 'Viagra' or 'Lotto Winner'?"
• Technical Check: "Was the email sent from a server known to distribute spam?"

4. Bayesian Filtering (sa-learn): The Smart Learner 

The Bayesian filter is SpamAssassin’s artificial intelligence engine that learns what spam and good mail look like specifically for our users. This is where the tool called sa-learn comes in.

• Its Job: This filter uses a branch of statistics called Bayes' theorem to calculate the probability that an email is spam based on the words and patterns it contains. It doesn't rely on pre-set rules; it learns from the emails it sees.
• How it Learns:
• When you mark an email as Spam (or Junk), you are feeding that example into the sa-learn system. It looks at the words and patterns in that message—like the phrase "click here to unsubscribe" or an unusual combination of colors—and increases the "spam probability" for those features.
• When you move a legitimate email out of your spam folder and into your inbox (marking it as Not Spam or Ham), it learns that those specific words and patterns are safe.

The power of Bayesian filtering is that it adapts to new spam techniques faster than manually updated rules. It makes our spam defense smarter over time, customizing the filter to the unique mail that our community receives.

Did You Know?

You can view how an email was scored by looking at the message source or 'header.'  The exact process depends on the email program you uses. A google search will tell you how to view this for your client.
This is a snippet from the relevant part of the header:

X-Virus-Scanned: amavis at starpoint.net
X-Spam-Flag: NO
X-Spam-Score: -1.083
X-Spam-Level:
X-Spam-Status: No, score=-1.083 required=6.2 tests=[BAYES_00=-1.9,
 DKIM_SIGNED=0.001, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1,
 HEADER_FROM_DIFFERENT_DOMAINS=0.001, HTML_FONT_LOW_CONTRAST=0.001,
 HTML_MESSAGE=0.001, LONG_HEX_URI=1, LOTS_OF_MONEY=0.001,
 MIME_HTML_MOSTLY=0.1, RCVD_IN_DNSWL_NONE=-0.0001,
 RCVD_IN_VALIDITY_RPBL_BLOCKED=0.001, RCVD_IN_VALIDITY_SAFE_BLOCKED=0.001,
 SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_KAM_HTML_FONT_INVALID=0.01]

In Summary

1. RBLs refuse connections from known spam sources.
2. Amavisd handles the mail and makes sure it gets scanned.
3. SpamAssassin runs hundreds of rule-based tests and assigns a score.
4. Bayesian Filtering (via sa-learn) adds a layer of intelligence by learning what is spam and what is safe mail based on content and user feedback.

This four-part system works together to give you the most accurate and up-to-date protection from spam.