A High Performance Image-Spam Filtering System

This paper proposes a three-layer image-spam filtering system. The system filters the image spam by analyzing both the mail header and image. The first layer of the system only analyzes the mail header and its processing time is very fast; the second and third layers analyze the high-level feature and low-level feature of images respectively. The experiment result shows that the first layer identifies about 93.7% of image mails. Thus, most image mails will be analyzed and identified by layer 1; and the proposed system reduces the average processing time for each e-mail greatly. The experiment result also shows that the accuracy rate of the system is about 94%.