Blockchain-based Land Registration and Ownership Management System with AI-Assisted Document Verification

Abstract

Land registration in Nepal relies on a paper-intensive
process requiring repeated visits to government offices, with
limited transparency and a high risk of document tampering. This
paper presents a blockchain-based land registration and ownership
management ecosystem that digitizes the workflow while preserving
the legally mandated physical and biometric verification of
the Land Revenue Office. Citizens submit applications, complete
KYC, and upload documents online, while verification officers
review claims with an AI-assisted tamper-detection pipeline
combining optical character recognition (OCR) and classical
machine learning. Only SHA-256 digests of documents and
opaque property identifiers are recorded on the Ethereum Sepolia
blockchain, preserving privacy, while personal data and legal
documents remain in PostgreSQL. A state-machine workflow
enforces every step before a new Lalpurja certificate is issued.
On a synthetic dataset of 2,000 Lalpurja documents, the model
achieves a ROC-AUC of 0.854 and a recall of 70% at a 10%
false-positive rate. Measured API latencies confirm the practical
feasibility of the proposed approach.