Smart Data Governance: A Hybrid Deep-Machine Learning Model for Real-Time Privacy and Compliance Automation

View at publisher Summarise with VIRAT
Authors: Ravindra Putchakayala, Rohit Yallavula, B. Jegajothi
Cited by: 0
References: 15

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The emergence of IoT, IIoMT and smart-city infrastructures has resulted in data streams which are enormous, heterogeneous and need real-time control yet maintain privacy, security and regulatory compliance. Traditional centralized analytics solutions have high predictive performance, but are finding it harder and harder to remain relevant over time as data-sovereignty limits, high-privacy concerns, and lack of transparent auditability emerge. To overcome these shortcomings, this paper presents BC-FedHVAE-GRU, a blockchain-based federated hybrid deeplearning architecture that can be used to enable privacyconscious, compliant smart-data governance automation. The suggested system combines Hierarchical Variational Autoencoder (HVAE) to abstract privacy-sensitive features, a sequence modelling module that is based on GRU and that allows the representation of dynamic behavior and characterizing anomalies, and a federated learning (FL) architecture that allows learning without transferring raw data across different nodes. It includes a lightweight blockchain layer, which stores model changes immutably and enforces access permissions and offers audit traces that are complianceready based on smart contracts. Extensive experiments with IoT-like sensor datasets prove that BC-FedHVAE-GRU provides competitive methods of detection and has important privacy and governance benefits. The F1-scores of the centralized baselines were$0.942,0.859$and 0.707 with MLP, GRU, and CNN+GRU, respectively. The proposed model achieved 0.889 accuracy and 0.864 F1-score under the federated conditions exceeding all the sequence based baselines and close to centralized performance regardless of decentralized training restrictions. Moreover, BC-FedHVAE-GRU reached ROCAUC 0.957 and PR-AUC 0.943 but federated optimization increased F1 -score to 0.856 after ten rounds. The interpretability of the framework, provided by SHAP and an autonomous policy of governance implemented by a DQN, support the transparency of the framework and its independent compliance. In general, BC-FedHVAE-GRU will provide a unified, privacy-sensitive and audit-capable framework of efficient smart-data management in real-time, which is suitable to the regulatory and operational demands of next-generation distributed settings.

Fields of Study & Subjects

Privacy-Preserving Technologies in Data Blockchain Technology Applications and Security Network Security and Intrusion Detection

Publication details

Journal / source
International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS)
Pages
1–9
Publication date
8 January 2026
Language
EN