A Filter System Of Image SPAM Email

With the popularization of the Internet, e-mail has become an important tool which helps people to communicate with each other. Because of the sudden huge profits, however, the spam is becoming a more and more serious problem. The average size of the spam is 50K, which is ten times as big as a normal one. As the size and the amount of spam are bigger and bigger, most of the business e-mail servers are in danger of being crushed. Therefore, how to deal with spam becomes an emergency task in the fields of the Internet and information security.Inspired by the keyword-based text filter, this paper proposes an image filter which detects the spam image by matching with user-specified image content. We make a study on spam feature recognition. The corresponding feature of each image is taken after the collection and classification process. The detection of each spam is carried out under the analysis result of the image attachment. The proposed filter extensible for implementing other kind of spam, such as filtering text spam.This paper proposes a prototype system of image spam filter. At first, a user-specified image, which represents a series of spam, is snapped from an e-mail and all of the snapped images compose a "black list" of spam image. Then, the image of each incoming e-mail will be taken to search for a matching candidate in the "black list". At last, the e-mail will be judged as spam is the matching candidate exists. SIFT (Scale Invariant Feature Transform) algorithm is applied in image matching process. The whole process is carried in the background, so it has merely effect to the user side and the network.The proposed method supplies the user with more ways to specify the interesting area of spam image, which makes the spam filter more effective and free. On the other hand, the proposed method avoids the analysis on text and deals with the text as an image. In this way, the noise and transformation added on the image will have little impact to the filter. The experimental results demonstrated that the proposed method is effective.