Convolutional Neural Network Based Detection Approach of Undesirable SMS (Short Message Service) in the Cameroonian Context

With the low cost of mobile phones, SMS(Short Message Service) and the advent of communications software (whatsapp, telegram, etc.), many people communicate easily. Many commercial companies, entities or individuals also use SMS to send out advertisements, unwanted messages containing links or malicious content that may violate customer privacy. This raises the question of how to help users to avoid being trapped? Several works have been proposed to counter these unwanted messages (SMS SPAM). Most of these are only tested on messages written in English, are not adaptable to messages from bilingual countries and are based on datasets dating back to 2012. Since Cameroon is a bilingual country, in order to set up a model taking into account SMS written in French, English or mixed language, we create a Cameroonian SMS spam dataset. The aim of this work is to propose a convolutional neural networks(CNN) based model for the processing of the Spam SMS dataset in the Cameroonian context to classify a given SMS as SPAM or HAM. The results obtained after experimentation in terms of accuracy, precision, recall and F1-score gave respectively 99.59%, 98.3%, 98%, 98.1%.