Analysis and comparison of fax spam detection algorithms

Spam detection is one of the important problems in these days. Many spam detection methods were proposed, but fax spam detection is not popular. It not easy to directly use existing content-based spam detection methods for fax documents because the documents are processed as image rather than text. In this paper, we propose a fax spam detection framework which is based on keyword patterns by using an Optical Character Recognition (OCR) technique. To demonstrate how effective the proposed framework is, we analyze and compare three fax spam detection algorithms (rule based method, SVM based method, and naïve Bayesian based method) with 219 normal and 212 spam documents. Our recommendation is to use naïve Bayesian based method which is capable of achieving an accuracy of 92.49%.