Preserving privacy in edge computing

Rahman, M 2019, Preserving privacy in edge computing, Doctor of Philosophy (PhD), Science, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Rahman.pdf Thesis application/pdf 4.75MB
Title Preserving privacy in edge computing
Author(s) Rahman, M
Year 2019
Abstract Edge computing or fog computing enables realtime services to smart application users by storing data and services at the edge of the networks. Edge devices in the edge computing handle data storage and service provisioning. Therefore, edge computing has become a  new norm for several delay-sensitive smart applications such as automated vehicles, ambient-assisted living, emergency response services, precision agriculture, and smart electricity grids. Despite having great potential, privacy threats are the main barriers to the success of edge computing. Attackers can leak private or sensitive information of data owners and modify service-related data for hampering service provisioning in edge computing-based smart applications. This research takes privacy issues of heterogeneous smart application data into account that are stored in edge data centers. From there, this study focuses on the development of privacy-preserving models for user-generated smart application data in edge computing and edge service-related data, such as Quality-of-Service (QoS) data, for ensuring unbiased service provisioning.

We begin with developing privacy-preserving techniques for user data generated by smart applications using steganography that is one of the data hiding techniques. In steganography, user sensitive information is hidden within nonsensitive information of data before outsourcing smart application data, and stego data are produced for storing in the edge data center. A steganography approach must be reversible or lossless to be useful in privacy-preserving techniques. In this research, we focus on numerical (sensor data) and textual (DNA sequence and text) data steganography. Existing steganography approaches for numerical data are irreversible. Hence, we introduce a lossless or reversible numerical data steganography approach using Error Correcting Codes (ECC). Modern lossless steganography approaches for text data steganography are mainly application-specific and lacks imperceptibility, and DNA steganography requires reference DNA sequence for the reconstruction of the original DNA sequence. Therefore, we present the first blind and lossless DNA sequence steganography approach based on the nucleotide substitution method in this study. In addition, a text steganography method is proposed that using invisible character and compression based encoding for ensuring reversibility and higher imperceptibility.  Different experiments are conducted to demonstrate the justification of our proposed methods in these studies.

The searching capability of the stored stego data is challenged in the edge data center without disclosing sensitive information. We present a privacy-preserving search framework for stego data on the edge data center that includes two methods. In the first method, we present a keyword-based privacy-preserving search method that allows a user to send a search query as a hash string. However, this method does not support the range query. Therefore, we develop a range search method on stego data using an order-preserving encryption (OPE) scheme. In both cases, the search service provider retrieves corresponding stego data without revealing any sensitive information. Several experiments are conducted for evaluating the performance of the framework.

Finally, we present a privacy-preserving service computation framework using Fully Homomorphic Encryption (FHE) based cryptosystem for ensuring the service provider's privacy during service selection and composition. Our contributions are two folds. First, we introduce a privacy-preserving service selection model based on encrypted Quality-of-Service (QoS) values of edge services for ensuring privacy. QoS values are encrypted using FHE. A distributed computation model for service selection using MapReduce is designed for improving efficiency. Second, we develop a composition model for edge services based on the functional relationship among edge services for optimizing the service selection process. Various experiments are performed in both centralized and distributed computing environments to evaluate the performance of the proposed framework using a synthetic QoS dataset.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Science
Subjects Networking and Communications
Coding and Information Theory
Keyword(s) Privacy in Edge Computing
Lossless steganography
Lossless biosignal steganography
Lossless DNA steganography
Lossless text steganography
Privacy preserving search
Search over stego data
Privacy preserving edge service selection
Privacy preserving edge service composition
Range search over stego data
Version Filter Type
Access Statistics: 128 Abstract Views, 41 File Downloads  -  Detailed Statistics
Created: Thu, 23 Apr 2020, 15:08:50 EST by Adam Rivett
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us