Based On Multi-feature Fusion Spam Filtering System

With the rapid development of the Internet, e-mail has been come into our lives, and become an important tool in our daily communication. However, the ensuing spam also increase, especially in recent years flooded the image-spam which give us a lot of inconvenience. So, how to fast and effective filter spam, especially the image-spam which has become an important problem in the field of Internet.This paper based on the existing mature text spam filtering technology, focus on the research and implementation of image-spam filter. Present the design and implementation of a combination filter which based on multi-feature fusion. By extract various spam picture feature, every feature is trained respectively to get their own single feature filter, then a combination of various single-feature filters to achieve the combination-filter which may optimize filtering effect of image-spam and can be convenient to add new features or remove original features. In this paper, combined with the matured text filtering technology, we also filter spam text which come which image-spam.The main work of this paper includes the following aspects:1. First introduced the background of spam, the definition of the emerging image-spam and its impact and detection difficult.2. Summary of the popular spam filtering technology, particularly focusing on the image-spam filtering methods, and then we analyze common spam images classification algorithms and their advantages and disadvantages and scope of their application.3. Analysis of the differences between the spam image and non-spam image in the various inspects include color, texture, shape and so on, propose features have each individual training, get a single feature-filter, and then by a single feature-filter combination to get a multi-feature spam filter.4. A simple implementation of Chinese text filter which based on Naive Bayes classification Algorithm, mainly used to filter the text contained in the image spam filter.5. Design and implement multiple single-feature spam image filters, then fuse various single-feature spam image filters, constitute a combination of multi-feature image spam filter.6. Implement a simple mail receive client which integrates text filter and imagefilter. 7. The detailed testing of each single characteristic image filter, combination ofimage filter, text filter.