Bayesian Filter

Bayesian filter, also known as Bayesian spam filtering, is a statistic-based process of email filtering that is employed to identify either wanted communication or as spam. This filter correlates the use of typical words with spam and non-spam emails. These commonalities are then used to calculate the probability that an email is or isn’t spam. The Bayesian filter can tailor itself to the responses of an individual email user, which makes it a powerful tool that yields a low rate of false positives.

This filtering technique first appeared in 1996 as a first generation of Bayesian email classification used to sort emails into folders but not to necessarily filter or reject emails as spam. In the years that followed, it was improved to be responsive to a users preferences. In essence, over time, the filter “learns” to distinguish spam or non-spam email. For example, if a user has marked an email newsletter that they did not subscribe to as spam, then a new trigger is added to the filter. This email or other types of communications will likely contain words that are common to future messages. These might include an email address, IP, or other data included in the body of the message or the header. The Bayesian filter will assign a higher probability to this type of content based on a user’s actions. In contrast, if a user commonly receives emails from the same address, or actively reads an email newsletter, the Bayesian filter will parse the message to the inbox automatically.