Research on a Naive Bayesian Based Short Message Filtering System

There are many researches on junk e-mail filtering but few on junk SMS filtering. This paper introduces a distributed SMS filtering system which is applicable on mobile network. This system has self-learning and knowledge updating capability and it can find junk SMS sender with a proper high credibility. The main algorithm used in this system is the naive Bayesian classification algorithm. Some attributes such as the length of the SMS and rules found by statistics are added to attribute set, and experiments show that it results a better performance than the traditional word based Bayesian approach. This paper also gives an approach to rank the suspicious SMS senders on their probabilities to be real junk SMS senders according to some measures