Integrating Siamese Neural Networks with Blockchain for Secure Identity Verification in Nigerian Educational Institutions

Abstract

Identity verification remains a critical challenge in Nigerian educational institutions, particularly in high-stakes processes such as examinations, admissions, and certification. This study proposes a hybrid identity verification framework that integrates Siamese Neural Networks (SNNs) for biometric face verification with blockchain-based smart contracts for secure and tamper-resistant identity management. The SNN learns discriminative facial embeddings using a transfer-learning backbone, while a Solidity smart contract deployed on an Ethereum test network (Ganache) stores cryptographic hashes of verified embeddings to ensure immutability, auditability, and decentralized access control. Experimental evaluation was conducted using structured train, validation, and test splits of student facial identities, followed by an extended verification protocol involving 128 genuine–impostor pairs. Performance was assessed using Receiver Operating Characteristic (ROC) analysis, threshold optimization, and bootstrap confidence intervals, yielding an Area Under the Curve (AUC) of 1.000 with a 95% confidence interval of (1.000, 1.000), and an optimal threshold producing an F1-score of 1.000 on the evaluation set. An ablation study comparing Siamese distance learning with cosine similarity demonstrated comparable separability within the current dataset, while robustness testing under minor image perturbations confirmed stability of the learned embeddings. Despite these promising results, the dataset size and diversity remain limited, and therefore the reported performance should be interpreted as a proof-of-concept rather than full generalization. The blockchain component was successfully deployed and tested with seven registered student identities, demonstrating secure on-chain storage and verification of biometric hashes, though real-world scalability and latency require further investigation. Overall, the proposed AI–blockchain framework demonstrates the feasibility of combining biometric deep learning with decentralized infrastructure to enhance identity integrity in educational systems and provides a foundation for secure, transparent, and auditable student identity management in Nigerian higher education, with future work focused on validating scalability, fairness, and performance on larger and more diverse datasets.