2023, Vol. 6, Issue 2

author(s)
Caroline I Samson-Onuorah

abstract
Financial fraud has evolved alongside the digitalisation of banking, exploiting high-velocity payments, platform integration, and cross-border data flows that challenge traditional rule-based controls. Modern banks now operate complex ecosystems spanning core banking systems, real-time payment rails, open banking interfaces, and cloud analytics, creating both unprecedented visibility and new attack surfaces. Against this backdrop, data-driven fraud detection has emerged as a strategic capability that combines machine learning, continuous transaction monitoring, and institutional governance to manage risk at scale. This paper situates fraud detection within a broader socio-technical context, examining how advances in data engineering, artificial intelligence, and regulatory expectations reshape financial crime control. It reviews the progression from static rules to adaptive models, highlighting supervised, unsupervised, and hybrid learning approaches used to identify anomalous behaviour, evolving fraud patterns, and coordinated attacks in real time. Particular attention is given to feature engineering, model explainability, and drift management as prerequisites for operational reliability and regulatory acceptance. The analysis then narrows to an integrated framework for modern banking platforms, where transaction monitoring pipelines, machine learning models, and governance mechanisms operate as a unified system. The framework emphasises lifecycle controls covering data provenance, model validation, auditability, and human oversight, ensuring that automated detection supports accountability rather than obscuring it. By aligning technical performance with risk governance, the paper demonstrates how banks can reduce fraud losses, improve detection latency, and meet supervisory expectations without sacrificing customer experience. The contribution provides a structured reference for practitioners and policymakers seeking resilient, transparent, and scalable fraud detection architectures in increasingly digital banking environments. It also outlines future research directions and implementation considerations for diverse institutional maturity levels globally applicable.