Towards Performance Evaluation of Oblivious Data Processing Emulated with Partially Homomorphic Encryption Schemes

摘要

Oblivious data processing has a potential to make an extremely significant impact on the whole landscape of data processing, specifically in scenarios that could be classified as big data. Recent developments in the area of fully homomorphic cryptography make oblivious data processing much more realistic. However, performance implications could render these approaches uneconomical. Other approaches such as a Download-and-Compute approach or a hybrid approach, though more cumbersome, could appear more efficient. This work presents early results of a project aimed at a comprehensive performance analysis of various techniques for oblivious data processing.