Research Of Semantic-Based Advertising Image Spam Filter Technology

Electronic mail (E-mail) has become an important way for communication and is successful in internet applications. In recent years, spam filters have been widely used based on message content, combined with the theory of machine learning, text classification and information filtering technology, but these methods have some limitations such as the image format. Therefore, as the growing of image spam, how to identify and filter junk e-mail image has become an urgent problem for IT sector and e-mail service providers.First, the problem of spam filtering status has been analyzed in this paper, mainly including the definition of spam, hazards, the current mainstream of spam filtering technologies and their strengths and weaknesses, etc. The key issues, in view of advertisement image characteristic, underlying vision feature extraction method can be improved, then we establish sample database of advertising image underlying multiple vision feature.Secondly, with the new feature of image spam and the analysis of the behavior and sending spam e-mail content, a semantic feature based on image similarity detection Image spam filtering methods has been proposed based on the use of large quantities of spam, content repeatedly sending and high degree of similar characteristics. This method will be achieved by detecting spam e-mail image and similarity of images. A variety of low-level visual features of the e-mail image will be extracted and mapped to high-level semantic features in order to determine whether the message image advertising junk mail image. The related issues and the key technology of this also be discussed, including image similarity measure and the semantic feature map.Finally, we apply the proposed method to junk mail filtration system which we also experiment on. The experiment results demonstrate that advertising junk mail filtration system based on image semantics similarity examination is highly accurate.