Comprehensive Performance Analysis of Homomorphic Cryptosystems for Practical Data Processing
This work addresses the scarcity of practical homomorphic data processing tools for secure cloud data management, but it is incremental as it focuses on performance analysis rather than new methods.
The study analyzed the performance of five homomorphic cryptosystems, finding that partially homomorphic ones like Paillier and ElGamal are practical for certain applications today, while fully homomorphic systems like HElib and SEAL are not yet ready for practical use.
Oblivious data processing has been an on and off topic for the last decade or so. It provides great opportunities for secure data management and processing, especially in the cloud. At the same time, modern computing resources seem to be affordable enough to allow for practical use of homomorphic cryptography. Yet, the availability of products that offer practical homomorphic data processing is extremely scarce. As part of a project aimed at developing a practical homomorphic data management platform, we have conducted an extensive study of homomorphic cryptosystems' performance, the results of which are presented in this work. For this work we chose the following five cryptosystems: fully homomorphic HElib and SEAL, somewhat fully homomorphic PyAono, and partially homomorphic Paillier and ElGamal. In the discussion of the aggregated results, we suggest that partially homomorphic cryptosystems could be used today in certain practical applications, whereas time has not yet come for the fully homomorphic ones.