Spamfire Auto-Training

Spamfire uses multiple filters to identify good and spam messages in your email. By using multiple filters, Spamfire can achieve very high accuracy. Spamfire can also train itself automatically by using the results of one filter to train another filter.

An example of auto-training

Suppose you receive a brand new spam message, one that you have never received before. The Senders filters will not recognize it because you have not added this sender to your senders list. The Rule filters will not recognize it because you have not created a Rule filter for this kind of email. And the Bayesian filter will not recognize it, because it has never been trained with the words in this email message.

But Spamfire comes with a Filter Update Subscription that lets you benefit from the experience of other Spamfire users. When a Spamfire user reports a spam message to us, we can create a Remote Filter for that spammer. You get access to these Remote Filters automatically.

In this example, Spamfire recognizes the message because of a Remote Filter. This prevents the spam message from making it to your email program. If it weren't for the Remote filter, you would have received the spam message. But Spamfire not only intercepts the spam message, it automatically trains the Bayesian filter to recognize similar messages in the future.

Thus, the Bayesian filter becomes more intelligent because of auto-training from the Remote Filters. As time goes on, the Bayesian filter will be able to identify new forms of spam, without your having to do anything to train it.

Manual training

Of course, you can also train Spamfire manually. Any time Spamfire makes a mistake, you should correct the mistake to improve future accuracy.