Loading [a11y]/accessibility-menu.js
BlockAIM: A Neural Network-Based Intelligent Middleware For Large-Scale IoT Data Placement Decisions | IEEE Journals & Magazine | IEEE Xplore

BlockAIM: A Neural Network-Based Intelligent Middleware For Large-Scale IoT Data Placement Decisions


Abstract:

Current Internet of Things (IoT) infrastructures rely on cloud storage however, relying on a single cloud provider puts limitations on the IoT applications and Service Le...Show More

Abstract:

Current Internet of Things (IoT) infrastructures rely on cloud storage however, relying on a single cloud provider puts limitations on the IoT applications and Service Level Agreement (SLA) requirements. Recently, multiple decentralized storage solutions (e.g., based on blockchains) have entered the market with distinct architecture, Quality of Service (QoS) parameters and at lower price compared to the cloud storage. In this work, we introduce BAM: a neural network-based middleware designed for intelligent selection of storage technology for IoT applications. We first propose a blockchain-based data placement protocol and theoretically model a decision optimization problem, which jointly considers cloud, multi-cloud and decentralized storage technologies to select the appropriate medium to store large-scale IoT data, while ensuring data integrity, traceability, auditability and decision verifiability. We then propose a neural network-based maintenance reconfiguration, which aims to optimize the computational complexity of the middleware design along with the blockchain transaction and storage overhead by learning and predicting the applications parameters. We also propose the aggregation rate feedback functionality in our design and model it as a linear optimization problem to improve data quality and precision. Finally, we provide a reference implementation and perform extensive experiments, which demonstrate the effectiveness of the proposed design.
Published in: IEEE Transactions on Mobile Computing ( Volume: 22, Issue: 1, 01 January 2023)
Page(s): 84 - 99
Date of Publication: 07 April 2021

ISSN Information:


Contact IEEE to Subscribe

References

References is not available for this document.