AI-Powered Fraud Detection in Serverless Financial Systems: Overcoming Latency and Scaling Bottlenecks

This paper presented a concrete, mathematically grounded implementation of Artificial Intelligence integration within a domain-specific, fully serverless architecture. It vividly demonstrates the practical, highly efficient application of ma-chine learning for real-time fraud detection in the financial sector. The proposed platform illustrated exactly how modern event-driven cloud infrastructure can be seamlessly aug-mented with AI-based inference to radically improve trans-action security while simultaneously slashing computational costs. The system proved that historically cited serverless limitations—namely statelessness and cold initialization de-lays—can be effectively engineered away through optimized model deployment pipelines and ultra-fast, NoSQL feature-store databases. Evaluation results confirm definitive improvements across inference speed, predictive precision, and architectural cost efficiency. The serverless paradigm, as implemented herein, meaningfully surpasses the technological utility and economic viability of traditional, constantly provisioned monolithic ar-chitectures in handling highly volatile financial workloads. Future research vectors will focus aggressively on incorpo-rating federated learning techniques, allowing decentralized, localized serverless nodes (e.g., edge computing via AWS Lambda@Edge) to update the global model in real-time with-out ever centralizing sensitive user data. Additionally, we will explore the deployment of quantized Graph Neural Networks (GNNs) within serverless memory limits to better identify and dismantle highly coordinated, multi-node synthetic fraud rings. The frameworks, mathematical models, and design patterns presented in this paper offer a highly replicable, robust foundation for software engineers and data scientists tasked with building the next generation of intelligent, hyperscalable financial security applications.

[1]     A. Dal Pozzolo, O. Caelen, R. A. Johnson, and G. Bontempi, “Cali-brating Probability with Undersampling for Unbalanced Classification,” in IEEE Symposium Series on Computational Intelligence, 2015, pp. 159-166. [2]     E. Jonas et al., “Cloud Programming Simplified: A Berkeley View on Serverless Computing,” arXiv preprint arXiv:1902.03383, 2019. [3]     Y. Sahin, S. Bulkan, and P. Duman, “A Cost-Sensitive Decision Tree Approach for Fraud Detection,” Expert Systems with Applications, vol. 40, no. 15, pp. 5916-5923, 2013. [4]     J. Wang, et al., “Machine Learning for Financial Fraud Detection: A Review,” Journal of Network and Computer Applications, vol. 182, 2021. [5]     P. Singh and S. Suri, “Serverless Architecture: A Paradigm Shift in Cloud Computing,” IEEE International Conference on Cloud Comput-ing, 2020. [6]     Amazon Web Services, “AWS Lambda Developer Guide,” [Online]. Available: https://docs.aws.amazon.com/lambda. [Accessed: May 2026]. [7]    K. J. Leonard and W. J. Wood, “Rule-based expert systems for fraud detection,” Journal of Retail Banking, vol. 17, no. 1, pp. 45-50, 1995. [8]    R. J. Bolton and D. J. Hand, “Statistical fraud detection: A review,” Statistical Science, pp. 235-244, 2002. [9]    Y. Xia, et al., “A Boosted Decision Tree Approach Using Bayesian Hyper-Parameter Optimization for Credit Card Fraud Detection,” IEEE Access, vol. 6, pp. 25684-25692, 2018.