Fully Homomorphic Encryption (FHE) is certainly one of the most exciting and promising developments in modern Cryptography, with the potential to revolutionize all contemporary forms of electronic communication via true end-to-end encryption including in processing stages. Yet, the speed of homomorphic operations—1,000x to 1,000,000x slower than their plaintext counterparts—currently prevents this potential from being realized in most commercially-viable applications.
Software improvements will only go part way to bridging this gap - we will show how the Fourier-optics hardware developed by Optalysys will be a major part of the solution in enabling FHE to reach its full potential.
After a very brief introduction to the core operations of FHE and what makes them slow, we will provide a short review of the physical principles behind Fourier optics and a high-level explanation of how it can, in principle, eliminate the main bottleneck in FHE. We will then show how our technology leverages this theoretical potential to develop a practical solution in a time frame of a few years. Finally, we will present results from numerical experiments using a modified version of the Concrete library showing the speed-up this technology can provide in the next three years. We will also take this opportunity to show how easy Concrete, the Google FHE transpiler, and other libraries make it for companies outside the traditional cryptography community to work with FHE, another important point for future wide adoption.
Florent Michel is an Applications Engineer at Optalysys.