Cybersecurity in Digital Banking: Advanced Fraud Detection Mechanisms for Financial Transactions

Abstract

Fraud detection in digital banking is increasingly important as more transactions are conducted online, and fraudsters become more sophisticated in their methods. In this paper, present a successful approach to fraud detection using a Recurrent Neural Network (RNN) on the IEEE-CIS fraud detection dataset. The proposed data preprocessing ensured the data was high-quality and suitable for the model to learn from. The proposed RNN model can learn temporal relationships in transaction data, which is vital for fraud detection. The proposed model showed great performance in the experiment, with an accuracy (acc), precision (prec) and F1-score (F1) of 94.34% and a recall (rec) of 94.38%. The model correctly classified both legal and illegal transactions, as shown by the confusion matrix, and the accuracy and loss curves indicated stable model training and low overfitting. Comparing the proposed RNN model with more traditional machine learning (ML) models, such as Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest (RF), showed that the proposed model performed better. The findings demonstrate that the proposed RNN is effective for digital banking fraud detection.