Spam Reader - Bayesian Spam Filter for Microsoft Outlook Home Features Download Buy Pro Screenshots Support About us Download Area Version 3.5 Click here to see the difference between Free and Pro versions. It's a straight forward yet comprehensive spam filter for Outlook combining Bayesian filtering with smart white lists and works with all accounts... Heinz Tschabitscher, email.about.com I'm very pleased with Spam Reader - it does a better job than built-in Outlook junk mail filter. I receive much spam and your product is a big help for me. David Solomon, www.solsem.com Our Users Spam Reader is great! Its learning algorithm has gotten to the point where it seems flawless. Very little spam gets through, and no valid material is kicked out. Mark Sarakis Bayesian approach and spam filtering Spam Reader 3.5 The main problem for anti-spam software is the correct spam recognition. Various techniques were developed and tried. But spammers quickly make them obsolete using even newer tricks. certain event in an experiment based on the results of this experiment. In the other words, suppose you are making Given the results of the experiment you can use the Bayes formula to estimate the probabilities of some events happening inside the black box. How does the Bayes approach work? To realize the Bayesian approach it is necessary to have an initial message database that contains both spam and non-spam messages (approximately equally). The base is used for learning the spam filter: for every word appearing in those messages the number of entries in spam messages and in non-spam ones is counted. This number helps to estimate how often the word appears in spam and non-spam and to calculate the corresponding probability. As a result, for both parts of the message database there are lists that contain the used words with corresponding probabilities. The lists are used to estimate the probability that a new message is spam. When a new message is received the spam filter divides it into words and finds the words in both lists. Using the probabilities for single words from the lists and applying the Bayes formula the program estimates the probability for the whole message to be either spam or non-spam. The main advantages of a Bayesian spam filter. Obviously, the effectiveness of spam filtering depends very much on the message database used for learning. So, to make filtering the most effective for a particular user it is necessary that the spam filter could learn on the message database of this user and automatically make adjustments when filtering is incorrect. Both this options are realized in Spam Reader and result in the following advantages: Possibility to customize spam filtering immediately after the installation; Possibility to correct filtering mistakes in order to avoid them in the future. Who was the first to use the Bayesian approach? The inventor and the main promoter of the idea to use the Bayesian approach in spam filtering software is Paul Graham, an American essayist, a piece of e-mail contains spam. Since then this approach has become very popular and inspired most current spam filters that use it as a part of the spam detection algorithm.