Privacy-preserving Platforms for Computation on Hybrid Clouds

Abstract

This thesis addresses the issues of data security and privacy in cloud computing, using a light-weight approach of segregating computation under the emerging hybrid cloud setting. More specifically, the thesis studies how to partition and schedule computations on hybrid clouds to achieve both security and efficiency.

This thesis first considers computation in the MapReduce paradigm and proposes a new model for MapReduce that supports tagging of sensitive data. It presents several ?scheduling modes? to re-arrange computation on hybrid clouds for improved performance, and a generic security framework for analyzing and comparing information leakage by different schedulers. Next, the focus shifts to computation in the stream processing paradigm in the domain of video surveillance. The thesis considers the scheduling problem, which is modeled as an integer programming problem and solved using a heuristic. Evaluation on Amazon clouds demonstrate that, privacy-preserving computation on hybrid clouds can be made efficient, cost-effective and automatic.