A Multilingual Web-Based SMS Phishing Detection System Tailored to the Nigerian Threat Landscape

Abstract

With mobile communications becoming part of Nigerian daily life, SMS phishing attacks are taking a huge turn for the worse, targeting unsuspecting users with fraudulent messages that trick them into providing personal and financial information. This study proposes a user-centric SMS phishing detection system tailored to the Nigerian context to design and implement an intelligent and user-friendly solution capable of accurately detecting phishing SMS messages and to raise user awareness using multilingual messages and easy feedback channels. The four classification algorithms trained and tested were Support Vector Machine (SVM), Random Forest, Multinomial Naive Bayes and Logistic Regression, whose performance metrics were accuracy, precision, recall, F1-score and Area Under the ROC Curve (AUC-ROC). Cross-validation was employed to ensure reliability and generalisability of the model. The result obtained from the models showed that the Support Vector Machine model performed best with the highest overall accuracy and good generalisation error in the classification of high-dimensional text data; thus, it was selected as the final model to be deployed. To enhance access by multiple user groups, the system was further enhanced by the provision of a web-based interface, several Nigerian language supports, visual warnings, and audio prompts. In conclusion, the developed system correctly identifies phishing messages and provides an easy way to reduce fraud on mobile devices in Nigeria. Future research should focus on expanding the dataset with more native language samples, adding deep learning models to improve multilingual understanding and partnering with telecommunication firms to make real-time fraud prevention and extensive use.