Home // International Journal On Advances in Security, volume 13, numbers 3 and 4, 2020 // View article

DFASC: Distributed Framework for Analytics Security in the Cloud

Authors:
Mamadou Diallo
Christopher Graves
Michael August
Kevin Groarke
Michael Holstrom
Megan Kline

Keywords: Homomorphic Encryption; Cloud Computing; Privacy; Data Analytics; Data Sharing

Abstract:
Processing big data requires advanced technologies that can extract useful information from large scale data to support decision making. These advanced technologies are currently being offered in the form of analytic tools hosted in the cloud and are being developed using different techniques such as artificial intelligence, machine learning, data mining, and statistical analysis. However, these tools are not very secure since the data they operate on must be in plaintext in the cloud, thereby leaving the data vulnerable to both insider and outsider attacks. To address these security issues when running data analytics in the cloud, we propose DFASC, a Distributed Framework for Analytics Security in the Cloud. At the core of the framework is homomorphic encryption (HE), which enables operations to be performed directly on encrypted data without using the private decryption key. Using HE, DFASC can distribute homomorphically encrypted data and analytics into the nodes of a distributed system and allow the analytics to operate on the encrypted data in each node. As a framework, DFASC provides mechanisms to enable the incorporation of HE libraries and data processing algorithms into the framework, which can then be used to implement analytic tools. A fundamental challenge with HE is its performance overhead due to the computationally intensive HE operations. This challenge of accelerating individual HE operations needs to be solved before secure big data processing in the cloud can be made practical. The distribution of the analytics not only improves the performance of the underlying analytic algorithms, it also helps to speed up the underlying HE operations. To enable the sharing of the encrypted data between parties in the cloud, DFASC incorporates a cryptographic key management infrastructure. To analyze feasibility of the framework, it was extended to implement a system that classifies images using a Neural Network algorithm. The experimental results show performance improvement of the system, including in HE operations, as the number of nodes in the cluster is increased.

Pages: 149 to 161

Copyright: Copyright (c) to authors, 2020. Used with permission.

Publication date: December 30, 2020

Published in: journal

ISSN: 1942-2636