A Rapid Image Spam Recognition Method Based on Content Feature Fusion Decision

Abstract

As a variation of spam, image spam can evade the text based filter simply and effectively. Based on the characteristic that this kind of mail is usually generated in manner of template and the image content under the same template are highly similar, this paper proposes a rapid method for identifying image spam by fusing content features. The layout features are first adopted to localize the spam speedily, and then the wavelet transform based content feature fusion decision algorithm is used to determine the spam accurately. The experimental results demonstrate that the proposed method is fast, accurate and practical.