SMS Phishing Detection with Hybrid CNN-GRU

Abstract

Smishing is a type of attack that allows access to personal and account data by activating various emotions, especially curiosity, of mobile device users through a link placed in an SMS (Short Message Service) text. Smishing can lead to unauthorized access to individuals' photographs, banking details, email contents, and a variety of other personal information. SMS text data has a complex non-linear structure. Because of this structure, detecting smishing with traditional approaches becomes a difficult task. Due to its nature, identifying smishing using conventional methods poses a challenging endeavor. To overcome these challenges, researchers have embraced deep learning approaches and achieved successful outcomes. Within the scope of this study, various pre-processes were applied to the smishing data set using NLP (Natural Language Processing) techniques to detect smishing. Then, the Hybrid CNN-GRU (Convolutional Neural Network - Gated Recurrent Unit) model was employed and the data were trained. As a result of application, an accuracy rate of 99.96% was achieved with the proposed method. The hybrid CNN-GRU method aims to prevent the attack attempt by detecting differences between normal user behavior and anomalies of the attacker's content. In line with the findings obtained as a result of the study, smishing attacks could be clearly identified. © 2024 IEEE.