AUTOENCODER-BASED CLOUD FRAMEWORK FOR DIGITAL BANKING: A DEEP LEARNING APPROACH TO FRAUD DETECTION, RISK ANALYSIS, AND DATA SECURITY

This paper presents an Autoencoder-Driven Cloud Framework for Digital Banking, leveraging both AI-driven deep learning techniques and cloud computing to enhance fraud detection, risk analysis, and data security in digital banking systems.The framework employs autoencoders to model normal customer transaction patterns and identify anomalies indicative of fraudulent activities.By integrating a cloud-based architecture, the framework ensures scalability and adaptability to high transaction volumes, while the AI model detects discrepancies with high precision.The cloud infrastructure also supports real-time processing and distributed data storage, optimizing both performance and resource utilization.The proposed framework was evaluated using the Fraud Detection Dataset from Kaggle, demonstrating outstanding performance across both AI and cloud metrics.The AI model achieved an accuracy of 98.6%, an F1-score of 0.94, a precision of 0.95, and a recall of 0.93, outperforming traditional fraud detection techniques.In terms of cloud performance, the system achieved latency of 150 ms, ensuring rapid transaction processing and detection, and a throughput of 1.2 million transactions per minute, demonstrating the system's ability to handle large volumes of data.Additionally, the system showcased scalability with a 45% improvement in transaction processing speed as the number of users increased, 99.99% availability, ensuring high reliability and minimal downtime, and cost efficiency, achieving a 30% reduction in cloud resource costs due to optimized resource allocation.These results highlight the potential of combining AI with cloud computing to optimize fraud detection, strengthen data security, and improve risk management in digital banking systems.