Research Based On P2p Technology, The Collaborative Anti-spam Filters And Realization

With the prevalence of Internet, E-mail has become one of the most important communication tools. However, the spam flooding has become one of the biggest problems harassing Internet. Based on the analysis of some kinds of current main spam filtering technologys, this article proposed a novel hybrid cooperation method for spam filtering. For absorbing mutual advantages, this method has combined client filtering method with server collaborative filtering method. The client side classified mail by means of Bayes algorithm on the principle of "only judge known mail", so known mail can be classified as legitimate mail and spam and unknown suspicious mail was sent to server for disposal. While the mail server side buildup a collaborative filtering network by structured peer-to-peer technology, accumulated spam knowledge and filtered various spam and their mutations sharing with other nodes.The main contents are shown as follows.Firstly, Junk Mail recognition mechanism: flooding delivering is one of the distinct characters of spam. In this paper, we recognised spam with counting the number of mails delivered in mail server side. When the number reached the threshold, this mail would be recognized as a spam.At the same time, this result would be published into P2P network providing with other members.Secondly, building interest community: Considering users' individuation, different users had different views about the same mail. While people who have similar interests had the same opinion. Based on local correlation principle, this paper has given an algorithm to create a interest community in Chord network.Thirdly, mail router algorithm: This paper designed a router finger structure according to mail filter and interest correlation based on Chord network and had given router process at the same time.Fourthly, mail final evaluation algorithm: At the initial phase of system running, the local filter may have a lower accuracy due to the small size of local knowledge database. And with the database grows, The final decision on whether a message is spam or not depends on two factors: local rating score and global rating score in the spam likelihood of the message.Finally, the summary of the whole work is given, while some of unsolved problems in the thesis and the prospect of the further study are indicated.