Design And Implementation Of The Intelligent Mail-head Analysis Mechanisms Based On LDAP

With the development of Internet, Email has become one of the most basic and the most universal information commutation methods for people in Internet era because of rapidity and cheapness. Email brings us convenience; however, it is abused by some people at the same time. As a result, Internet is often flooded with spam. Spam has affected not only the normal Internet application but also people's normal work and study. It has brought great loss to Email user and Internet service provider. Thus, how to filter spam effectively is an important direction in the current Internet application research.In this thesis, the primary technology of spam filtering are summed up and analyzed and the conclusion that spam prevention is mainly depending on technique method in recent years can be gotton. It is a valid method to judge and filter spam in MTA (Mail Transfer Agent). Email has many characteristics; it is composed of mail-head and mail-body. There is much information of the mail-head can be used as important foundation to judge spam. And according to part of the mail-head information, the source of spam can also be tracked, thus the spamer can be imposed sanctions.LDAP directory service is designed to optimize read-intensive operation. The response capability of server can be increased greatly by using LDAP. Thus, OpenLdap database server is chosen in this thesis. Mail-head information of Email samples is gotton, vectors of mail-head information are generated using Vector Space Model and analysed using intelligent algorithms. Considering the sizes of samples, classification precision, off-line training time when the number of sort increased and sensitive words, four intelligent mail-head analysis mechanisms are designed, respectively based on scalable decision tree algorithm, variable precision rough set decision tree algorithm, class incremental decision tree algorithm and quick BP (Back Propagation) neural network algorithm. According to the four algorithms, mail-head vectors are analyzed, features are selected, and filtering rules are gotton. These rules can be directly used at MTA to block the spam.To verify the feasibility and validity of the four mail-head information analysis mechanisms, these mechanisms are implemented, and performances are tested in the following sides:the time of getting rules, the capability of finding spam, the accuracy of judging spam, and the rate of taking ham for spam. After that, the performances of these mechanisms are compared, and the testing results have shown that these mechanisms have satisfying performances. Thus, the mechanisms designed by this paper are both feasible and effective in spam filtering application.