Phishing Detection in Arabic SMS Messages using Natural Language Processing

Mobile phone integration into daily life has elevated Short Message Service (SMS) to a crucial tool for communication. Users receive text messages from banks, electronic government services, businesses, and payment services to verify their identities. Which makes them a source of manipulation to gain access to personal data. This study proposes a technique that detects Arabic phishing messages using natural language processing and a random forest classifier. The performance of the random forest classifier is compared with other machine learning algorithms, namely, K-Nearest Neighbors (KNN), AdaBoost, and Logistic Regression. According to all evaluation matrices, the random forest classifier has outperformed other classifiers. The model was trained with 638 phishing messages and 4844 legitimate ones. The experimental outcomes indicate that the proposed approach has obtained an accuracy of 98.66%, 99.10% precision, 98.23% recall, and 98.67% F1 score.