ML-Enhanced EduTrust-Chain: A Random Forest-Based Pre-Validation Framework for Fraud Detection in Academic Credential Management

Abstract

Blockchain is proving to be an effective technology for secure and tamper-resistant academic credential management. However ,in a traditional blockchain credential system, all transactions submitted to the blockchain, including fraudulent ones, are processed by the consensus mechanism which leading to unnecessary computational overhead. To overcome this limitation, this paper proposes an ML-Enhanced EduTrust-Chain framework that combines an education-specific blockchain architecture with a Random Forest-based pre-validation mechanism. The proposed framework would filter credential transactions from the mempool before they enter to the blockchain consensus process and reducing the chance of a fraudulent transaction being added to the blockchain.To evaluate the proposed framework, a synthetic dataset was generated, consisting of 50,000 academic credential transactions,including 40,000 legitimate while 10,000 were fraudulent transactions covering four education-related fraud scenarios,The experimental results show that the Random Forest classifier achieved an accuracy of 99.37% and an F1-score of 98.42%. In addition, the pre-validation mechanism reduced the number of transactions entering the blockchain consensus process by 19.91%. This reduce the consensus workload and optimizes resource utilization. Overall, the results show that the integrating machine learning with blockchain improves the efficiency, scalability, and reliability of academic credential management systems, while preserving the trust model of EduTrust-Chain.