REAL-TIME FRAUD DETECTION IN HEADLESS COMMERCE USING FEDERATED LEARNING
DOI:
https://doi.org/10.46121/pspc.54.2.11Keywords:
: Federated Learning, Fraud Detection, Headless Commerce, Composable Architecture, Differential Privacy, Secure Aggregation, Real-Time Inference, Api Security, Fedavg, Machine Learning PrivacyAbstract
The present-day digital retail world has adopted headless commerce architectures which separate their online display systems from their backend business functions through their API-first system structure. The new system design creates difficulties for fraud detection because transaction data gets dispersed through various API connections and merchant sites, and e-commerce platforms maintain strict data protection standards which block organizations from collecting customer usage patterns that standard fraud detection methods need. The paper presents a real-time fraud detection system for headless commerce environments which uses a federated learning framework to operate. The system enables merchants to train their models together while keeping their transaction information protected, and it achieves inference times under 100 milliseconds through its edge-based model deployment system. The research results show that federated machine learning systems enable organizations to protect their private data and meet regulatory requirements while maintaining security through their advanced privacy features.
