Predictive System for Fraud Prevention in E-Commerce Transactions in Peru Using Facial Recognition and Random Forest

Alva-Cordova F.; Reyes-Suarez L.; Wong L. · Applications in Software Engineering - Proceedings of the 14th International Conference on Software Process Improvement, CIMPS 2025 (2025) · DOI: 10.1109/cimps69356.2025.11419706

Resumen

The exponential growth of e-commerce has significantly increased digital fraud, exposing new vulnerabilities such as identity theft and payment manipulation that traditional systems often fail to mitigate effectively. To address this challenge, a dual-layer predictive system for fraud prevention in e-commerce transactions in Peru is proposed. This hybrid approach integrates a classification model based on the Random Forest algorithm with facial recognition technology for user authentication. The proposal was developed in four phases: dataset creation, classification model training, model integration as web services, and deployment of a system with biometric authentication. The system was evaluated through experimentation with 20 real users in a simulated online store environment, where 200 transactions were processed using a functional mobile application. The results demonstrated a significant quantitative improvement, achieving an F1-Score of 93.55%, while reducing undetected fraud by 66.7% and erroneously blocked legitimate transactions by 50%. The system was successfully validated and showed high user acceptance, with participants valuing the added security and usability.

Instituciones: Universidad Peruana de Ciencias Aplicadas