A new SMS spam detection method using both Content-Based and non Content-Based features

Abstract

SMS spamming is an activity of sending ‘unwanted messages’ through text messaging or other communication services; normally using mobile phones. Nowadays there are many methods for SMS spam detection, ranging from the list-based, statistical algorithm, IP-based and using machine learning. However, an optimum method for SMS spam detection is difficult to find due to issues of SMS length, battery and memory performances. Hoping to minimize the aforementioned problems, this paper introduces another detection variance that is based on common characters used when sending SMS (i.e. numbers and symbols), SMS length and keywords. To verify our work, the proposed features were stipulated into five different algorithms and then, tested with three different datasets for their ability to detect spam. From the conduct of experiments, it can be suggested that these three features are reasonable to be used for detecting SMS spam as it produced positive results. In the future, it is anticipated that the proposed algorithm will perform better when combined with machine learning techniques.