Design and Deployment of AI-Driven Fraud Detection Models in Cloud-Based Banking Ecosystems

In this paper, we propose AI-based fraud detection model for Cloud Banking Ecosystem. Using advanced machine learning techniques such as Transformer-based architectures, the model detects fraudulent transactions accurately. We show that our model significantly outperforms traditional methods over important metrics like detection speed, precision, recall, and F1-score. It delves into the implications of adversarial attacks on the robustness of the model and also explores the scalability of the model with increasing transaction volume. Moreover, the feature importance analysis also helps to understand the decisionmaking process, contributing to the model interpretability of the model in compliance with the applicable regulation. The findings underscore the effectiveness of the model for fraud detection while identifying further improvements needed in attack/defines machinery, especially robustness against adversarial attacks.