The statistical trainable classification algorithm has unique characteristics. It learns your private correspondence, extracts the distinguishing features from messages of different classes and uses that information for effective analysis of other incoming messages. The algorithm improves its classification quality continuously with each new learned message. The messages from whitelisted senders are not filtered by other methods, and are recognized as non-spam at once. The whitelist is updated on learning automatically. The filtering rules are recognizing spam and ham by specific combinations of symbols in a body and headers of the email messages.