Research Of Image Spam Detection Technology Based On Classification And Similarity

Along with the development of the networks and information technology, the email has become the most important mean for people to communication. However the flooding of spam has led to tremendous inconvenience to the people's life. Especially starting from 2005, the spam makers embed spam information into graphical images with the technology of noise adding,fuzzification and so on, thus the image spam comes into being. How to detect image spam accurately and efficiently is an urgent problem.This paper analyses the background,development actualities and significance of the studies on image spam systematically, studies the key technology of image spam deeply. The paper put forth the image spam detecting method based on the classification and near-duplication. The following are the main innovation of this paper:1. Analyzing the various characteristics of image in spam detecting, putting forward the method of using local invariant features to describe the distribution of spam information.2. Putting forth the spam detecting method based on classification since the various image spam construction ways. The method is constructed by two combined filters, the first is the filter based on the noise features extracting in images, the other is the combined filter of color,texture and shape characteristic. The experiments show this combined filter has very high recall.3. To the shortcoming of the traditional image spam detecting system, we use SURF to extract the local invariant features, and then train the classifier based on the GMM to detect image spam. The experiments in the paper show:comparing with other detecting systems, this method of image spam detecting has much better filter results.