Research On Anti-Spam Technology Based On Behavior Patterns Classification

With the popularity of Internet, e-mail with convenient, speedy, low-cost features becomes one of the main communication tools on the Internet and the most important one of the most popular applications in modern society. At the same time, spam has become increasingly rampant, and occupies a large number of limited storage, computing and network resources on the Internet, and reduces efficiency in the use of the network. Users spent a lot of processing time.Therefore, the study of efficient spam filtering technology is of great significance. For the research on anti-spam technology based on behavior pattern classification, papers give a multi-level spam-filtering model based on behavior recognition and its application to the enterprise e-mail filter projects. The behavior-based spam filtering technology avoids technical limitations of the traditional content-based filters, and creates the behavior-based recognition model to recognize spam by mining and using the various key characteristics in the process of sending and receiving e-mail. The technology has achieved that the criterion has nothing to do with the content of the e-mail, has nothing to do with the language, so as to further ensure the stability of behavior-based filtering and users'privacy.The main thesis work and contributions:1. By mining the behavior features in the course of e-mail conversation, spam identification model based on the MTA is proposed. It does not need to receive the whole e-mail; spam filtering will be advanced to the conversation stage.2. By mining the normal user behavior features, spam recognition model based on the MDA is proposed. Compare to changing of e-mail content, behavior features is almost fixed, so it is better than the other filtering methods based on the MDA, and it has better stability.3. By mining the normal sending behavior, spam-filtering model based on the sending MUA is proposed. An e-mail is judged on client before being sent. It saves network resources and inhibits bad senders sending spam based on e-mail client and inhibits controlled accounts become accomplices of spammers in unknown circumstances.4. By analyzing normal user receiving behavior characteristics, spam filtering model based on the receiving MUA is proposed. Users can set up their respective rules, and then make further identification on the e-mail client.5. The anti-spam technology based on behavior patterns classification is applied to the e-mail subsystem in enterprise information system and a comprehensive information management platform is formed.