Children aged 10–18 in Oman face a growing cyber threat landscape—cyberbullying, online grooming, phishing, and harmful content—yet no AI-based protection framework exists that addresses the Omani Arabic dialect context. This paper makes two contributions. First, it presents empirical evidence of the online threat environment facing Omani children, drawn from structured surveys administered to 75 participants (40 children, 35 parents), revealing widespread threat exposure and a statistically significant parental awareness gap (χ² = 7.84, p = 0.020). Second, it proposes a conceptual AI-based protection framework leveraging AraBERT fine-tuned on an annotated Omani Arabic corpus, and defines the evaluation methodology—using Precision, Recall, and F1-Score against keyword-filtering baselines—for planned future deployment. Target benchmark thresholds are derived from comparable Arabic NLP studies. The paper concludes with evidence-based policy recommendations for the MTCIT and the Ministry of Education.
Al-Shaibi et al. (Sun,) studied this question.