Research On Anti-spam Technology Based On Shannon Entropy

At present network security events have increased rapidly. The spread of spams is one of these bad incidents. The traditional anti-spam technologies are mainly based on the filtering of mail contents. And they are devided into based on the statistics and based on the rules. But these methods have certain limitations, as they can't work well on effectiveness and speedness simultaneously. In this paper we introduce Shannon entropy to solve this problem. We build a compositive filter system to distinguish spams in front of mail servers, based on mails' behavior parameters and contents. It raises accuracy rate and reduces false rate.This paper mainly includes three aspects below:Firstly it improves naive bayes algorithm by adopting multilevel attributes related, and enhances its practicability and applicability. On the other hand it analyses bayes network in the polytree condition. Using parallel processing it computes the input messages and gets a good performance.Secondly, according to spams'characters, the paper uses Shannon entropy which is also called information entropy. As spam having been sent regularity with lower randomness, the paper calculates some characters parameters of mails, and filters spams on the basis of the results.At last, the experiment part integrates modules to build a comprehensive email filtering system. After mail participle and principal component analysis, the system calculates entropy of mail parameters and uses the bayes classification. From the experiment results, the system has got a expectant effect and improved the efficiency. So it has a practical value.