Edge enabled digital payment analytics ecosystem machine learning based suspicious activity identification uncertainty scoring model

Abstract

The rapid expansion of edge-enabled digital payment ecosystems has significantly transformed global financial transactions by enabling real-time, low-latency, and distributed payment processing. However, this evolution has also intensified the exposure of digital financial infrastructures to sophisticated cyber threats, including fraudulent transactions, ransomware-driven extortion, anonymized illicit transfers, and blockchain-based laundering activities. Existing rule-based fraud detection systems are insufficient in addressing the dynamic, adaptive, and obfuscated nature of modern financial crimes.

This research proposes an Edge Enabled Digital Payment Analytics Ecosystem (EEDPAE) integrated with a machine learning-based suspicious activity identification framework enhanced by an uncertainty scoring model. The proposed system leverages edge computing architecture to enable localized transaction analytics while maintaining global intelligence synchronization through centralized learning nodes. Machine learning classifiers are employed to detect anomalous transaction behaviors across digital payment networks, while uncertainty quantification mechanisms provide probabilistic confidence scores for each detection outcome.

The framework is conceptually grounded in blockchain forensics and cryptocurrency behavior analysis literature, including studies on Bitcoin transaction graph analysis, anonymization challenges, and ransomware payment tracking (Nakamoto, 2008; Meiklejohn et al., 2013; Reid & Harrigan, 2013). Additionally, insights from cybersecurity threat intelligence reports and empirical machine learning comparisons support the design of robust classification and anomaly detection models (Caruana & Niculescu-Mizil, 2006; Symantec Corporation, 2017).

A key contribution of this study is the integration of uncertainty scoring into edge-based fraud detection pipelines, allowing the system to distinguish between high-confidence suspicious activities and ambiguous transactions requiring further verification. Furthermore, the architecture aligns with cloud-assisted fintech intelligence paradigms that emphasize scalability, adaptive learning, and real-time risk assessment in financial ecosystems (Goyal et al., 2026).

Experimental synthesis from literature indicates that hybrid machine learning models combined with graph-based transaction analysis significantly improve detection accuracy in digital payment systems. The proposed framework enhances interpretability, reduces false positives, and strengthens resilience against evolving cyber-financial threats.