Arabic SMS Spam Detection

The rise of social engineering attacks in recent years has highlighted the critical need for effective spam detection systems. This study explores Arabic short message service (SMS) spam detection using machine learning techniques, addressing the linguistic and cultural challenges of Arabic spam messages. A dataset was carefully constructed to reflect real-world spam patterns in Saudi Arabia, including spam keywords, unsecured URLs, and phone number patterns. Several classifiers were evaluated, including Random Forest, Naïve Bayes, and Support Vector Machines (SVM), as well as a Combined Model (Random Forest + SVM). The Combined Model achieved a performance with overall accuracy (94%), with a precision of 0.94, recall of 0.93, and F1 score of 0.94 effectively balancing the strengths of both classifiers. This work underscores the importance of tailored feature engineering, threshold tuning, and data preprocessing in achieving high detection accuracy for Arabic SMS spam. These findings enable future research by establishing a basis for investigating real-time deployment and adaptation to evolving spam tactics.