Date of Award

9-4-2015

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Shucheng Yu

Abstract

The continuous development of cloud computing is increasingly attracting individuals and organizations to outsource their data to public cloud platforms. While enjoying advantages brought by cloud computing, cloud users also raise their concerns on data security with cloud computing in terms of integrity and privacy. This is because cloud users lose physical control of data and services after outsourcing them to cloud. In this dissertation, we identify security challenges with data services in cloud computing, and present a set of novel solutions that address security issues with data services outsourced to cloud. Specifically, we focus on ensuring the integrity and privacy of data storage and utilization in public cloud. The first part of this dissertation discusses the integrity assurance for data storage of cloud computing. Considering that cloud users may not have physical access to data stored on cloud, we introduce a series of remote data integrity auditing schemes. Specifically, we first design a public integrity auditing scheme that is suitable for archival or backup data stored on cloud. Then, we consider the integrity auditing for shared dynamic data stored on cloud, and propose an efficient integrity auditing scheme to support multi-user scenario. Our schemes can be applied to not only cloud storage applications, but also other cloud based applications such as software version control systems. The second part of this dissertation targets at privacy-preserving data utilization in public cloud. To protect data privacy, cloud users can encrypt their data with traditional encryption algorithms before storing them on cloud. However, these algorithms also prevent cloud servers from performing required data analytics. To utilize the power of public cloud computing while protecting the privacy of cloud users' data, we propose novel encryption schemes that allow cloud servers to perform desired operations over encrypted data. Specifically, we first present a lightweight secure image search scheme over encrypted data, which achieves similar search efficiency and accuracy as compared with search techniques over unprotected images. Next, we design a privacy-preserving neural network learning algorithm, which enables multiple parties to securely delegate their data to public cloud for learning without disclosing their data privacy.

Download

Share

COinS