Systematic Literature Review on CNN-LSTM Model for Smishing Detection in Hausa and English Messages

Abstract

Despite advancements in machine learning and cybersecurity, traditional rule-based and machine learning (ML) techniques struggle to keep pace with the continuously evolving tactics of cybercriminals. Deep Learning (DL) models such as hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) models, have demonstrated improved performance in smishing detection in a mixed mobile environment supporting Hausa and English messages. This work provides a systematic literature review (SLR) following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Synthesizing the 41 selected papers using the SLR approach by analyzing the DL - based smishing detection methods, identified previous research efforts, datasets type for models training, their effectiveness, limitations and operational challenges such as computational and algorithms complexities. This review provides a clearer understanding of smishing attacks, refinement of detection algorithms and discusses research gaps and future directions to address current challenges and improvement of smishing detection systems.