AI Against Smishing in Kenya: Culturally Adapted SMS Scam Detection for Digital Trust

Kenya’s mobile-financial ecosystem, driven by M-Pesa and a mobile penetration rate above 133%, has advanced commerce and financial inclusion but also exposed users to SMS-based scams exploiting linguistic diversity, cultural trust, and psychological manipulation. Existing fraud-reporting mechanisms such as keyword filtering and manual reporting, remain reactive and inadequate against these evolving threats. To fill this gap, this study develops a culturally adapted, machine learning–driven approach to SMS scam detection tailored to Kenya’s multilingual environment. Using a crowdsourced dataset of 738 SMS messages (427 scams, 311 legitimate), XGBoost achieved 83.8% accuracy with strong precision and F1-scores, while Logistic Regression offered superior recall (91.8%) in detecting fraudulent content. SHAP analysis revealed urgency cues, code-switching features, and social proof language as the most influential predictors of scams. The proposed model, designed for integration via APIs into USSD platforms, enables real-time detection and fosters user confidence through explainable alerts. This study provides a scalable, context-aware framework for fraud prevention in mobile-first economies and actionable insights for telecom operators, fintech platforms, and policymakers seeking to strengthen digital trust.