SMS Spam Filtering Technique Based on Artificial Immune System

The Short Message Service (SMS) have an important economic impact for end users and service providers. Spam is a serious universal problem that causes problems for almost all users. Several studies have been presented, including implementations of spam filters that prevent spam from reaching their destination. Naïve Bayesian algorithm is one of the most effective approaches used in filtering techniques. The computational power of smart phones are increasing, making increasingly possible to perform spam filtering at these devices as a mobile agent application, leading to better personalization and effectiveness. The challenge of filtering SMS spam is that the short messages often consist of few words composed of abbreviations and idioms. In this paper, we propose an anti-spam technique based on Artificial Immune System (AIS) for filtering SMS spam messages. The proposed technique utilizes a set of some features that can be used as inputs to spam detection model. The idea is to classify message using trained dataset that contains Phone Numbers, Spam Words, and Detectors. Our proposed technique utilizes a double collection of bulk SMS messages Spam and Ham in the training process. We state a set of stages that help us to build dataset such as tokenizer, stop word filter, and training process. Experimental results presented in this paper are based on iPhone Operating System (iOS). The results applied to the testing messages show that the proposed system can classify the SMS spam and ham with accurate compared with Naïve Bayesian algorithm.