AI-Driven Fraud Detection in IoT-Enabled Payment Ecosystems: Challenges, Hybrid Edge-Cloud Framework, and Emerging Trends

Financial fraud in IoT-enabled payment systems has emerged as a significant challenge due to the rapid adoption of digital transactions, real-time processing requirements, and the increasing sophistication of fraud schemes. Fraud detection involves identifying unauthorized, anomalous, or suspicious transactions to prevent monetary losses and safeguard user trust. Traditional detection approaches, such as rule-based systems and classical machine learning (ML) models like Logistic Regression (LR), Random Forest (RF), and XGBoost (XGB), provide a baseline capability for flagging simple or known patterns. However, these methods often struggle with temporal transaction dynamics, networked collusion, and previously unseen fraudulent behaviors, leading to high false positives or missed attacks. To address these limitations, a hybrid AI-driven framework is proposed, combining baseline classifiers with advanced models, including Long Short-Term Memory (LSTM) networks for capturing temporal spending patterns, Graph Neural Networks (GNN) for modelling interaccount relationships, and Autoencoders (AE) for unsupervised anomaly detection under label-scarce conditions. The framework leverages ensemble techniques to fuse model outputs, producing a robust risk score while maintaining low detection latency suitable for IoT environments. Evaluation on simulated mobile payment data demonstrates that the model achieves a precision of $96.8 \%$, recall of $94.5 \%$, F1-Score of $95.6 \%$, ROC-AUC of 0.98, and inference latency of 38 ms, outperforming individual baseline and advanced models. The results highlight the framework’s effectiveness in capturing complex and novel fraud patterns. Future trends include the integration of federated learning for privacy-preserving crossplatform training, adaptive thresholding for dynamic risk assessment, and edge-cloud collaboration for scalable real-time deployment. The proposed model provides a practical and extensible solution for enhancing security in IoT payment systems.