Research And Implementation Of Spam Filtering System Based On The Sender Abnormal Behavior Detection

With the developing of modern network technology, E-mail has become an very important tool for our fast-paced life, which is attributed to its particular technology advantage of fast communication and large amounts of information. But the spam mail also brings negative effects to the state and enterprises. Therefore, the study in the technical difficulties of spam mail has significant academic meaning and application value. Currently there are many achievements of research on the anti-spam technology. But these techniques only work out during a short period due to the rapid changing of spam techniques. The anti-spam technology based on behavior analysis not only proves the sender’s privacy, but also ensure the efficient filtration performance.After analyzing and summarizing up-to-date technique on spam filtering field, thesis designs and implements an anti-spam system based on sender abnormal behavior detection. The system according to mail client behavior characteristics in the process of SMTP session, finds the email sender abnormal behavior. When sending emails, system will record and save the change rule of characteristics which showed the same IP address corresponding to the domain and the number of link to server port, also finds email sender abnormal behavior, and realizes filtering spam on time dimension extension. Bring in the artificial neuron network algorithm to detect spams, and use neuron centered ANN presentation method with the purpose of improving system efficiency.Through a variety of indices of system performance evaluation, it shows good performance of Spam Filtering System Based on Sender Abnormal Behavior Detection. The result of test reduces the heavy burden of mail servers, also proves high precision rate and good recall rate at the same time.