Game Theoretic Analysis on Encrypted Data Deduplication in Cloud

Abstract

Deduplication is a technology for removing duplicated data by only storing one copy in the cloud. User concerns about security and privacy lead them to store sensitive data in an encrypted form. Existing encrypted cloud data deduplication (Here, referred to as "Deduplication") schemes can be classified into three types: server-controlled deduplication (S-DEDU), client-controlled deduplication (C-DEDU), and hybrid deduplication (H-DEDU), based on which stakeholder (Cloud Service Provider (CSP), data owner or combination of both) can control the deduplication. However, the deduplication schemes widely proposed in the literature are rarely applied in practice. Rational CSPs and data users are profit-driven and decide whether to adopt a deduplication scheme based on the benefits gained from it. Storing one copy for each data item narrows the relationship between data users and amplifies the impact of adverse behaviors, such as data disclosure and mismanagement in CSPs. In this dissertation, we apply game theory as the main methodology to capture the dynamic interaction of system stakeholders in the three types of deduplication schemes and to design incentive mechanisms for motivating all involved stakeholders to participate and cooperate. We build a non-cooperative game between CSPs and data holders in S-DEDU, in which all players choose whether to participate in deduplication. We propose a bounded discount-based incentive mechanism for promoting the willingness of data holders to participate in S-DEDU while at the same time guaranteeing the profits of CSPs. In the case of C-DEDU, we define a unified discount model and an individualized discount model and investigate their applicability in C-DEDU. We establish a non-cooperative game model between a data owner and a data holder, where each player chooses to follow or deviate from the deduplication scheme design. By analyzing the Nash Equilibrium of this game model, we identify a free-riding behavior in the unified discount model and a privacy violation problem in the individualized discount model, as well as innovate two solutions for promoting cooperation while at the same time preserving data privacy. Finally, in the case of H-DEDU, we refine the two-dimensional action sets in S-DEDU and C-DEDU by transforming them into uncountable ones, as well as build a Stackelberg game for studying the interaction among a CSP, a data owner, and data holders. We analyze the cooperative behaviors of all players for revealing the optimal strategies employed by the stakeholders and prove the existence of a Nash Equilibrium in the H-DEDU system.