Over the last decade, there has been rapid growth in cloud computing with cloud-based services commonly used for a variety of applications including healthcare, financial, transportation, and many others. While the cloud-based services offer a way to efficiently execute these applications, these cloud-based services also introduce privacy concerns. FHE is a promising solution to address these privacy concerns due to FHE’s ability to compute on encrypted data. FHE allows the cloud to operate on the data without having to decrypt it. Unfortunately, the large compute and memory requirements of FHE remain a serious barrier to its widespread adoption.
The goal of this tutorial is to provide an introduction to the key concepts of Fully Homomorphic Encryption (FHE)-based computing, discuss the challenges associated with architecting FHE-based computing systems, and then design FHE-based computing systems from the ground up. This tutorial will capture these limitations by first presenting the fundamentals of the operations for the Cheon, Kim, Kim, and Song (CKKS) scheme, one of the commonly used FHE schemes and then present an analysis of various commodity platforms when running applications that use CKKS FHE. This will be followed by a hands-on design of a hardware accelerator for CKKS FHE-based computing.
At the end of the tutorial, attendees will have an understanding of the challenges, opportunities, and pitfalls associated with designing CKKS FHE-based computing systems. We expect that this tutorial will allow many more computer architects to actively contribute to this growing field of FHE-based computing.
The target audience for this tutorial will be graduate students, faculty, and research scientists/engineers who are interested in learning about the basics of FHE, and about the process of designing FHE-based computing systems. It will be helpful if the attendees have a basic knowledge of linear algebra and finite-field arithmetic.