Blockchain Integrated AI for Secure Data Access and Sharing

Abstract

The rapid expansion of digital ecosystems across healthcare, finance, and critical infrastructure has made secure data 
access and sharing a paramount challenge of the decade. Centralised access control systems are vulnerable to single points 
of failure, insider threats, and lack tamper-proof audit mechanisms. This paper presents a Blockchain-Integrated AI (BAIA) 
framework that orchestrates permissioned Hyperledger Fabric smart contracts with a transformer-based behavioural 
anomaly detection engine for real-time threat identification and adaptive permission management. Each access request is 
evaluated against a continuously updated AI risk model trained on 18 months of enterprise access logs comprising 4.2 
million transactions across healthcare, financial, supply chain, and IoT domains. Smart contracts encode three tiers of 
access policy—static role-based rules, dynamic risk-adjusted escalations, and emergency lockdown—enforced 
deterministically on every network node. Experiments on a held-out three-month test set demonstrate that BAIA achieves 
97.3% anomaly detection accuracy with a false positive rate of 0.8%, processes access requests at 310 ms median latency, 
and reduces unauthorised access incidents by 89.4% compared to rule-based baselines. A federated training protocol 
allows participating organisations to improve the shared AI model without exposing raw access logs. The immutable ledger 
provides cryptographically verifiable audit trails for every access event, combining the auditability of blockchain with the 
adaptability of machine learning.