Zero-Trust Payment Infrastructures: A GenAI-Driven Threat Detection Mesh for Digital Wallet Ecosystems

he rapid proliferation of digital wallets, embedded payment services, and decentralized financial systems has amplified the threat landscape across real-time financial networks. Traditional rule-based fraud detection systems are no longer sufficient to counter increasingly complex, distributed, and adversarial cyber threats. This paper introduces a GenAI-powered cybersecurity mesh architecture designed to enforce Zero-Trust principles within digital payment ecosystems. The proposed system integrates transformer-driven threat intelligence with federated anomaly detection to continuously assess payment behaviors, device posture, session activity, and user identity signals. Key architectural innovations include a lightweight multi-layer mesh of edge nodes that process encrypted telemetry in real time, enabling sub-second threat scoring without compromising user privacy. The system leverages a continuous learning loop across geographies and wallets, combining supervised and unsupervised GenAI models to adapt to emerging threat signatures, with an explanation layer to support regulatory reviews and real-time dispute resolution. In simulations and a pilot deployment, the mesh achieved an ≈25% relative uplift in true-positive detection versus a strong gradient-boosted baseline (93.1% vs. 74.5% TPR) and ≈30% lower median inference latency (126 ms vs. 180 ms), while keeping FPR ≤ 2%. These results position GenAI-based cybersecurity meshes as a foundational capability for securing next-generation payment infrastructures.