Ambar Kusumaningsih, Ika Diyah Candra Arifah, Dian Anita Nuswantara, Insyirah Putikadea, Hapsari Shinta Citra Puspita Dewi, Erta
Traditional red flag approaches are increasingly inadequate for detecting modern financial fraud, which often involves complex and adaptive schemes. This study explores the potential of Artificial Intelligence (AI)specifically Machine Learning (ML), Deep Learning (DL), and Generative Adversarial Networks (GANs)-to enhance fraud detection beyond conventional methods. Guided by the Fraud Hexagon theory, a Systematic Literature Review (SLR) using the PRISMA framework was conducted across four databases: Scopus, IEEE Xplore, SpringerLink, and ScienceDirect. From 500 initial articles, 85 publications (2015-2025) were selected based on relevance, full-text availability, and thematic alignment. Python-assisted text mining and topic modeling revealed key trends and advances in AI-driven fraud detection. Results show that AI offers notable benefits in accuracy, adaptability, and real-time response. Additionally, Explainable AI (XAI) emerges as a critical component for transparency and compliance with regulatory standards. However, challenges remain, including algorithmic bias, ethical concerns, and data privacy issues. This study contributes a multidimensional analysis of AI applications in financial fraud detection, addressing both technical methodologies and ethical considerations-areas often overlooked in previous reviews. The findings suggest that hybrid AI models, when combined with strong governance frameworks, represent a promising direction for future audit and fraud prevention systems. © 2025 IEEE.
Universitas Negeri Surabaya, Department of Accounting, Surabaya, Indonesia; Universitas Negeri Surabaya, Department of Digital Business, Surabaya, Indonesia; Universitas Negeri Surabaya, Department of Sport Management, Surabaya, Indonesia
