Systematic review of the internet of things ecosystem for real-time detection of card-not-present (CNP) fraud

The rapid rise in card-not-present (CNP) transactions has increased fraud risks in digital commerce, highlighting the limitations of traditional fraud detection methods. The integration of Internet of Things (IoT) technologies introduces contextual data that can enhance real-time, context-aware fraud detection. However, it also presents challenges related to scalability, latency, privacy, and governance. This study systematically reviews IoT-enabled frameworks for real-time CNP fraud detection, synthesizing evidence from 23 peer-reviewed articles identified through a literature search covering the period 2015 to April 3, 2026, following the PRISMA 2020 guidelines. The analysis categorizes existing frameworks based on application scenarios, IoT architectural layers, performance evaluation, and distributed intelligence, revealing system-level trends and identifying research gaps. While many studies emphasize predictive effectiveness, performance reports mainly rely on accuracy metrics, often overlooking essential factors such as real-time responsiveness and privacy. Notably, federated and split learning are highlighted as significant enablers of scalable, privacy-preserving collaboration, although their coordination and communication trade-offs are often not quantified. Moreover, blockchain mechanisms can enhance auditability but require careful integration to address potential latency issues. This study provides several key contributions: (i) it introduces a scenario-aware synthesis that links IoT architectural placement with fraud detection requirements and evaluation practices; (ii) it provides explicit guidance for benchmarking that goes beyond accuracy to encompass latency, scalability/coordination cost, and privacy/trust indicators; and (iii) it presents a unified multi-layer IoT-CNP reference architecture that aligns with the CNP transaction lifecycle. Rather than proposing a new detection algorithm, this review enhances system-level integration and deployment readiness by consolidating fragmented evidence into a practical architecture and metrics blueprint for detecting real-time IoT-enabled payment fraud.