Research And Implementation Of Distributed Anti-Spam System Based On Bayesian

With the development of internet,email has been used widely in daily life.However, some people collect a large number of email addresses from internet for commercial purposes, sent advertisements, pornography, political mail to them, which greatly hurt the interests of ordinary email users and occupied ISP's network resources, system resources and storage resources. Control of spam has become a major issue of network research.This paper has analyzed the advantages and disadvantages of various anti-spam technologies and has done a deep research on Chinese word segmentation which restricts the performance of Bayesian filtering, have proposed Crossed n-gram algorithm which is suitable for Chinese mail filtering based-on Bayesian.The paper has studied the application of feature selection technology inmail filtering, found several better algorithms for feature selection, done a comprehensive experiment about how many features should be selected.Based on the research, the paper has proposed a new framework for distributed anti-spam system, and has realized it. The paper have added log information in system, in order to achieve the distributed store, query and share. Based on these, proposed a complete strategy of intelligent Bayesian learning and weights update. We have test the strategy through simulation experiment, and found it perform better than traditional Bayesian filtering.