Explainable multi-layer Arabic phishing detection system processes SMS, email, and messaging app text through preprocessing, rule-based and ML detection engines, then scores and classifies threats as legit, suspicious, or phishing. Preprocessing normalizes Arabic text via cleaning, tokenization, and stopword handling before feeding six parallel detection modules: rule-based keyword matching, URL analysis, social engineering pattern recognition, and Arabic linguistic analysis alongside ML classification. Risk scoring combines all module outputs with confidence estimation to produce a final label and human-readable explanation tied to specific detection indicators. Fork this diagram to customize detection rules, add new input sources, or integrate with your security operations platform.