Applying Fully Homomorphic Encryption: Practices and Problems

Abstract

Fully homomorphic encryption (FHE) has been regarded as the "holy grail" of cryptography for its versatility as a cryptographic primitive and wide range of potential applications. Since Gentry published the first theoretically feasible FHE design in 2008, there has been a lot of new discoveries and inventions in this particular field. New schemes significantly reduce the computational cost of FHE and make practical deployment within reach. As a result, FHE schemes have come off the paper and been explored and tested extensively in practice. However, FHE is made possible with many new problems and assumptions that are not yet well studied. In this thesis we present a comprehensive and intuitive overview of the current applied FHE landscape, from design to implementation, and draw attention to potential vulnerabilities both in theory and in practice. In more detail, we show how to use currently available FHE libraries for aggregation and select parameters to avoid weak FHE instances.