A BLOCKCHAIN IMMUTABILITY AND MACHINE LEARNING CERTIFICATE: FRAUD DETECTION FOR MITIGATING CREDENTIAL FRAUD

Abstract

Academic credential fraud is a critical systemic challenge in Nigerian higher education because conventional verification depends on the continued operation, record integrity, and responsiveness of issuing institutions. This study addresses the resulting research gap by developing a verification infrastructure that remains functional independently of institutional continuity. A blockchain-based academic certificate verification system was designed and implemented using two Solidity 0.8.20 smart contracts, InstitutionRegistry and CertificateRegistry, deployed on the Sepolia Ethereum testnet. Certificate documents were stored on the InterPlanetary File System (IPFS) through Pinata, while SHA-256 hashes and IPFS Content Identifiers were permanently anchored on-chain during issuance. A React 18 application provided role-specific interfaces for institutions, graduates, and third-party verifiers. Security evaluation demonstrated resistance across five attack categories, while 200-trial benchmarking recorded an average verification time of 87 ms at zero verifier cost. Polygon Layer 2 deployment reduced certificate issuance costs to approximately $0.018. A complementary machine-learning fraud-detection component used 1,000 structured certificate records, each containing 16 domain-specific features. Logistic Regression, Random Forest, and XGBoost were optimized using GridSearchCV with five-fold cross-validation and evaluated on held-out data. XGBoost achieved the highest accuracy (93.5%), followed by Random Forest (91.5%) and Logistic Regression (75.5%). Verification complexity, previous rejections, and days to verify were the most predictive features. The complete system and analytical pipeline are packaged in a reproducible Google Colab notebook.