A Sobel Edge Detection Algorithm Based System for Analyzing and Classifying Image Based Spam

Early spam mails were only text-based, however spammers have moved to more sophisticated spamming techniques that involve images now generally termed image based spam. In most image-based spam, the entire spam message, which could be sometimes text, is embedded in an image of any format. This type of spam emails creates another dimension to the spam filtering problem scenario. Extracting text from the image and filtering these text components is one method that has been used to deal with image spam with little success because Spammers modify their approaches to beat such filters even when such filters are based on Optical Character Recognition. In this work, we used employ the Sobel edge detection algorithm, which analyses a low level feature of an image as an alternative to the OCR only based filtering system. The low level feature resultant from the filtering activity is then used to calculate the global magnitude of the edge which aides in classifying the image as either spam or ham. Our system named WiSpaf can analyse images as well as photographic images and be able to tell them apart.