Predicting Suspicious Arabic X Accounts Using Stylometric Features

Abstract

Some users use the X platform to spread negativity, violence, and hatred. These users may initially conceal their true beliefs in order to gain users’ trust, making traditional content-based detection methods not helpful in identifying these accounts as suspicious (i.e., indicate whether an X account belongs to a terrorist). Stylometric features, which analyze writing styles, can reveal behavioral traits and hidden thoughts. In this paper, we propose a novel model that predicts whether an Arabic X account is suspicious (owned by a terrorist) using 85 stylometric features extracted from 1,500 accounts (750 suspicious, 750 non-suspicious). We utilized a variety of AI techniques, such as machine learning and deep learning, to evaluate our approach. We also used NLP techniques for preprocessing and feature extraction. Our results showed that the Random Forest (RF) model achieved the highest accuracy, reaching 98%. This approach can aid cybersecurity efforts by detecting suspicious accounts without relying on content analysis.