Decentralized Digital Identity: A Blockchain and Neural Network Approach
Role & Contribution
Co-Author · Lead Architect for N-SAM Biometric System
- Architected the N-SAM biometric pipeline for decentralized identity, integrating deep-learning facial embeddings with encrypted off-chain IPFS storage.
- Implemented and benchmarked FaceNet, FaceNet512, and ArcFace across LFW/BUPT datasets, achieving up to 97.65% accuracy and AUC ≈ 0.996.
- Designed the system’s feature extraction, preprocessing, and thresholding workflow, including ROC and cross-validation analysis.
Abstract
Current digital identity systems face significant challenges in privacy, security, user control, and performance. This research proposes a novel blockchain-powered approach that integrates neural network technologies to address fundamental limitations of centralized identity management. By leveraging decentralized architecture and biometric authentication, we present a transformative solution to digital identity verification that enhances user privacy, security, and autonomy while mitigating systemic risks in both centralized and decentralized frameworks. Empirical evaluation demonstrates authentication accuracy of up to 97.20% and average login latency of 1.069 seconds, validating the system’s effectiveness and responsiveness on standard consumer hardware.
Citation
Babu A, Balasubramanian KR, Singh A, Meenakshi RS and Natarajan Y. Decentralized Digital Identity: A Blockchain and Neural Network Approach. Premier Journal of Science 2025;15:100142