Predictive Model of Phishing Attacks Using Machine Learning for Fintech Companies in Peru

As Fintech companies undergo rapid growth, they have become increasingly susceptible to phishing threats, which pose a substantial risk to both user security and institutional integrity. This paper presents a proposed model that combines eXtreme Gradient Boosting (XGBoost), a highly effective gradient boosting algorithm known for its speed and accuracy, with Synthetic Minority Over-sampling Technique (SMOTE), a technique employed to mitigate class imbalance in phishing detection datasets. The innovative aspect of this combination lies in leveraging SMOTE’s ability to generate synthetic minority class samples, thereby balancing the dataset and enabling XGBoost to better learn the distinguishing features of phishing emails. The model was trained on a comprehensive dataset containing both legitimate and phishing emails, with the primary objective of accurately classifying and predicting phishing attempts to prevent potential breaches. The evaluation results indicate that the model exhibits a high degree of precision, recall, and overall detection rate. Furthermore, the results show that the model performs significantly better in handling imbalanced data through SMOTE. A significant finding of this study is the provision of a robust and scalable solution for Fintech companies, offering a proactive approach to enhancing cybersecurity and preventing financial fraud through email-based phishing attacks. This innovative approach has been proven to enhance cybersecurity resilience and protect fintech companies, thereby ensuring continued trust and stability in online financial services.