Open Access

REVIEW

A Systematic Review of Frameworks for the Detection and Prevention of Card-Not-Present (CNP) Fraud

Kwabena Owusu-Mensah^*, Edward Danso Ansong , Kofi Sarpong Adu-Manu, Winfred Yaokumah

Department of Computer Science, College of Basic and Applied Sciences, University of Ghana, Legon, Accra, P.O. Box LG 25, Ghana

* Corresponding Author: Kwabena Owusu-Mensah. Email:

Journal of Cyber Security 2026, 8, 33-92. https://doi.org/10.32604/jcs.2026.074265

Received 07 October 2025; Accepted 09 December 2025; Issue published 20 January 2026

Abstract

The rapid growth of digital payment systems and remote financial services has led to a significant increase in Card-Not-Present (CNP) fraud, which is now the primary source of card-related losses worldwide. Traditional rule-based fraud detection methods are becoming insufficient due to several challenges, including data imbalance, concept drift, privacy concerns, and limited interpretability. In response to these issues, a systematic review of twenty-four CNP fraud detection frameworks developed between 2014 and 2025 was conducted. This review aimed to identify the technologies, strategies, and design considerations necessary for adaptive solutions that align with evolving regulatory standards. The findings indicate a shift from static, supervised models to dynamic approaches, such as hybrid and federated architectures, which utilize advanced technologies like Graph Neural Networks (GNNs), blockchain auditing, and privacy-preserving learning. These modern frameworks demonstrate impressive performance metrics, achieving F1 scores between 0.85 and 0.99 and AUC values exceeding 0.93, while also complying with regulatory standards, including GDPR and PCI-DSS. The review identified six key design pillars essential for effective CNP fraud mitigation: scalable architecture, privacy-preserving governance, adaptive learning, interpretability, cost optimization, and integrated continuous evaluation. This study presents a design-centric framework that emphasizes scalability, ethical governance, and explainable intelligence. The review suggests that graph-enabled, federated, and self-optimizing frameworks represent the future of securing digital payment environments and enhancing CNP fraud detection.

---

Keywords

---