Digital Banking Security: Assessing Risks of Mobile Payment Systems Using Machine Learning

Abstract

Quick adoption of mobile payment systems through digitalization of banking industry has has also brought about potential security threats, such as unauthorized transactions, malware attacks, phishing schemes, SIM-swap fraud and exposure of digital identity. The objective of this study is to evaluate and forecast such security risks with a machine learning–based analysis model specifically built for mobile payment scenarios. A set of mobile payment comprehensive dataset about user behaviors, transaction patterns and risk indicators has been preprocessed and balanced by SMOTE algorithm due to class unbalance. Important features of risk classification were identified using the Select Best method based on the classify scoring function. Logistic Regression, Decision Tree, Random Forest and Support Vector Machine (SVM) machine learning models were implemented with accuracy, precision, recall, F1-score and ROC curve assessment. The testing results show that Random Forest model topped the ground with a testing accuracy of 86.27%, closely followed by precision 86.52% and F1- score 86.33% , proving its ability to well detect high-risk transactions and potential security threats from mobile payment system. The visualization of confusion matrix and ROC curve also reaffirmed the stable performance of the model in decreasing false positives and negatives, which is paramount for real digital banking deployment. These results indicate that feature selection, pre-processing, and ensemble learning methods can be effective for improving the prediction accuracy and interpretability of machine learning models. In conclusion, this research delivers a sound data-driven and scalable model for enhancing security of digital banking along with the risks involved in mobile payment systems.