Research On Junk Mail Filtering Model Based On POP3 In MUA~2

At present, the number of junk mail is more and more, so the research on mail filtering system is more and more important.In the location of junk mail filtering system, the way of MUA has the performance of low resource, low cost, easy realization, easy maintenance, etc. But at present the most of junk mail filtering systems are too close to mail client and are not independent and transparent to mail clients.In the method of mail filtering, the method based on statistic probability has the performance of high accuracy and low inaccuracy, so this method is widely used in junk mail filtering system. But there is a lack of universal junk mail filtering system which can use a variety of methods based on statistic probability.In the self-adaption of mail filtering, the method of machine learning is used to train and adjust the mail filtering model, so the performance of junk mail filtering system is better and better. But how to accept users'feedback from mail clients and adjust the mail filtering model by users'feedback is not solved at present.Based on the above background and problems, SAMFUF (Self-Adaptive Mail Filtering POP3 proxy based on Users'Feedback) is realized and proposed. The main work and innovations of thesis are as follows:(1) Realize great efficiency and stability of mail filtering by adopting thread pool technology and windows overlapped I/O sockets. It is the main work of the thesis and the follow three points are the ionnovations of the thesis.(2) Establish transparent connection between mail client and mail server by asynchronous connection in place of synchronous connection.(3) Realize the universality of email filtering system through separating Maximum Entropy, Bayes, SVM, etc. from email filtering system.(4) Realize the self-adaptive function by inserting the HTML label which includes the unique ID of an email, classified results and so on and providing the service of adjusting mail filtering model.Experiments show that SAMFUF has high efficiency, accuracy and universality in junk mail filtering.