Article Details

The rapid expansion of mobile connectivity in Nigeria has created a fertile ground for SMS phishing (smishing) attacks. Conventional detection methods often fail to adapt to the evolving tactics used by scammers, leaving the country's large mobile population vulnerable, Short Message Service (SMS) is still a vital communication tool in our daily life activities, even with the quick development of Internet protocol-based messaging services. This research focuses on the Nigerian context, analyzing local smishing campaigns to develop a more effective, tailored detection model. The study aims to enhance cyber security defenses specifically for Nigeria’s unique digital landscape. This thesis proposes building a domain-specific deep learning model for Nigeria. This system is designed to accurately classify SMS messages by directly tackling the issue of imbalanced data, and will be developed using ethically handled datasets to create a more robust cybersecurity defense. This study collects the dataset, which is the combination of SMS Smishing Collection from Kaggle and smishing messages from the Nigerian locally collected, ensuring relevance to the local context. The Pre-processing methods involved steps to manage missing and duplicated values, while checking label uniqueness, and performing text pre-processing and lemmatization, followed by label encoding. The dataset is balanced with Synthetic Minority Over Sampling Technique. Convolutional Neural Network, Long Short-Term Memory, and Attention Mechanism some of the deep learning classification models selected for their exceptional performance in text analysis. The models work well for detecting fake SMS messages evaluation matrix were used Accuracy, precision, recall, and F1 score. The result showed that the Hybrid (CNN+LSTM+ATTENTION) classifier achieving a superior accuracy of 99.3% compared to other models. This study highlights the practical implications of smishing detection, additionally, the research discusses potential future work, including the integration of transformer-based models, the handling of model drift, and addressing adversarial concerns in dynamic environments.