Securing supply chain transactions with deep learning based autoencoders for fraud detection and zero knowledge proof blockchain for data security

Purpose This research presents a privacy-preserving fraud detection framework that involves integrating an Autoencoder using a Deep Learning algorithm with Zero Knowledge Proof (ZKP) blockchain technology for secure transactions in the supply chain.Methodology The framework employs analysis of transaction, inventory, and shipment data that are extracted from the Kaggle supply chain management dataset in order to detect unusual transaction patterns as well as verify fraud detection results without any compromise on the privacy of the transaction.Key Findings The proposed DL-Autoencoder model yields an accuracy of 99.99%, precision of 99.87%, recall of 99.98%, and F1-score of 99.54%. These results are better than those obtained through Federated Learning, Anti-Adversarial Hierarchical Ensemble Learning with Multi-Layer Multi-Anomaly Detection (AHEAD), and Blockchain-based African Buffalo (BbAB). The use of ZKP-based verification facilitates privacy-preserving validation, while blockchain technology ensures immutability and traceability of verified fraud results.Policy Implication The framework provides a way for secure and scalable transaction of fraud detection in a supply chain environment.Original Contribution Therefore, the major contribution of the framework is the integration of autoencoder-based anomaly detection, ZKP-based verification, and blockchain-based record-keeping into supply chain fraud detection.