Arabic Phishing Guard implements a seven-layer explainable detection system combining rule-based analysis, linguistic feature extraction, and ML classification to identify phishing threats in SMS, email, and messaging app text. Input sources flow through preprocessing (text cleaning, Arabic normalization, tokenization) into a multi-layer detection engine evaluating suspicious keywords, URL patterns, social engineering tactics, and Arabic-specific linguistic markers. Results feed into scoring and decision layers that calculate risk scores and confidence estimates, producing final classifications (Legit, Suspicious, Phishing) with human-readable explanations and actionable recommendations (Safe, Warning, Block/Report). This architecture prioritizes explainability at every stage, enabling security teams to understand detection reasoning and tune thresholds for their threat landscape. Fork and customize this diagram on Diagrams.so to adapt detection rules, add provider-specific integrations, or extend the pipeline with additional language support.