Spam Mail Filtering System Based-on Network Processor

As the popularization of Internet, the efficient and cheep e-mail has become the substitute to the conventional papery mail. But at the same time, spam mails cause lots of problems. Content-based spam filtering technologies have become the mainstream anti-spam mail methods so far. Support vector machine (SVM), Bayes, windows and KNN are excellent ones of these technologies and they have advantages and disadvantages respectively. The content-based spam filtering methods consume lots of time on training sample mails and classifying testing mails.As the network bandwidth growing up, higher network data processing rate is required. Network processor is developed to satisfy the requirement. Multi-processors, special instruction set for network data processing and multi-hierarchies memory system make network processor be able to process network data from layer 2 to 7 efficiently and conveniently.This paper implements content-based spam mail filtering systems based- on SVM and Na(?)ve Bayes on network processor. The major work is as follows:1. Introducing protocols and standards relative to e-mails as while as the current anti-spam technologies; analyzing the characteristics of SVM and Bayes in spam mail fitering.2. Implementing a double classifier spam mail fitering system based-on SVM and Bayes. While maintaining the filtering accuracy of content-based method, the system takes advantage of the parallel processing ability of network processor to improve the filtering speed.3. Constructing the experiment environment and testing the performance of the system on Ling-Spam mail library; analyzing the results of the tests.