A Systematic Review of AI-Driven Banking Fraud Detection: Advances, Challenges, and Deployment-Ready Solutions (2024–2025)

Abstract

​Digital transformation has sharply increased fraud risks in banking, with global annual losses exceeding $50 billion. This systematic review analyzes 20 influential peer-reviewed studies from 2024 to 2025, selected through a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020–guided search from over 150 papers in IEEE Xplore Digital Library, SpringerLink, ScienceDirect, Scopus, and Google Scholar, using terms such as "AI banking fraud detection," "deep learning financial fraud," "graph neural network fraud," and "explainable AI banking." Included studies are published in reputable venues, report robust evaluations on real or large-scale transaction data, introduce meaningful innovations (e.g., novel architectures or privacy-preserving methods), and address deployment in banking environments. The reviewed works employ deep neural networks, graph-based models, hybrid ensembles, explainable artificial intelligence (XAI), and blockchain-inspired components, achieving detection accuracies of 93–96% on real transaction datasets. Nonetheless, major obstacles still limit large-scale adoption. We identify 10 recurring challenges; high computational costs, data quality problems, and limited model interpretability each appear in at least 30% of the studies. Our analysis compares the strengths, weaknesses, and contributions of existing approaches and highlights critical gaps in scalability, robustness, and standardization. We recommend wider use of lightweight models, broader application of federated learning, creation of shared benchmark datasets, and adoption of common evaluation protocols. Finally, we outline priority directions to move beyond a narrow focus on accuracy toward fraud detection solutions that are scalable, secure, transparent, and cost-effective, enabling more confident deployment of AI-based systems in real-world banking.

Received: 14 November 2025 | Revised: 14 February 2026 | Accepted: 14 July 2026

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

Author Contribution Statement

Hamid Banirostam: Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration. Elham Shamsinejad: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing.