Analysis on AI-based Techniques for Detection of Banking Frauds: Recent Trends, Challenges, and Future Directions

Banking fraud is a potential threat to both customers and financial organizations, risking financial stability. Traditional rule-based techniques are simple, but lack adaptability to evolving fraud, which leads to high false positive rates, causing inefficiencies, thereby necessitating the adoption of artificial intelligence (AI)-driven solutions. This study reviews 23 articles from 2023-2024 on recent AI advancements in fraud detection across three key themes, machine learning (ML), deep learning (DL), and hybrid models, evaluating their strengths, limitations, and performance. It highlights emerging technologies like Graph Neural Networks (GNNs), Federated Learning (FL), and Blockchain, addresses ethical AI concerns such as bias, fairness, and privacy, and identifies gaps in real-time detection, adversarial resilience, and scalability. Future research aims to integrate Explainable AI (XAI) and blockchain to develop transparent, secure, and scalable systems, fostering trust and compliance.