All Published Papers Search Service

Title

Detecting SMS Phishing Based on Arabic Text-Content Using Deep Learning

Author

Sadeem Alsufyani and Samah Alajmani

Vol. 25  No. 4  pp. 1-10

Abstract

SMS phishing is a type of cyberattack where fraudsters use text messages to deceive people into giving up confidential information, like their bank details or personal data. While a lot of research has gone into spotting these fake texts in languages like English, there's a real lack of studies on the topic in Arabic. This paper aims to bridge that gap by putting forward a model specifically designed to identify phishing in Arabic SMS messages. The proposed model involves several stages, beginning with the collection of a dataset containing Arabic SMS messages, followed by Arabic SMS dataset cleaning. The linguistic complexities of the Arabic language are then addressed through preprocessing in the textual content of the Arabic SMS messages, such as removing Arabic stop words, diacritics, punctuation, and other irrelevant elements. Since Arabic words can have multiple forms, they are reduced to their Arabic roots using a stemming process. Next, features are extracted using the TF-IDF technique, and the target classes are encoded. Finally, they are passed through the classification process. The model uses three various deep learning techniques: BiGRU, CNN, and GRU, to detect messages as either phishing or legitimate (ham). The study compares the performance of the three models for deep learning utilizing the four main criteria and demonstrates the outcomes that the BiGRU model, with an accuracy rate of 98.71%, outperformed the other models. The GRU model achieved an accuracy rate of 98.32%, and the CNN model achieved an accuracy rate of 97.86%. These outcomes demonstrate BiGRU's capability to Arabic phishing SMS messages detection.

Keywords

Phishing Threat; Arabic Phishing; Arabic SMS Text-content; Bidirectional GRU (BiGRU); Phone Phishing.

URL

http://paper.ijcsns.org/07_book/202504/20250401.pdf