Fully Homomorphic Encryption (FHE) is an exciting cryptographic technique for performing computations over encrypted data without first decrypting the data, meaning that the processor never learns anything about the data it is handling. User data remains completely secure from malicious actors, data leaks, and mishandled data regardless of where it is stored. FHE has been recognized as a powerful tool for privacy applications, e.g., machine learning, inventory/order matching in financial services and private cryptocurrencies.

One challenge to using FHE is that one needs to be an expert in lattice cryptography to set up a secure system, and even then it can take many months of development and testing to ensure that the encryption scheme is implemented properly and preserves data security.

In this talk we will discuss Sunscreen's new approach to FHE program creation where a developer writes regular C code that is transformed into a performant, secure FHE application able to process encrypted data without ever decrypting it. This is possible due to two technologies: (1) a novel virtual processor design that generates a representation of FHE programs in a parallelizable, portable, and compact way, and (2) a compiler that converts normal C code into the instruction set of the processor.

The FHE programs generated by this process are very close to the best possible performance possible from a human developed FHE program, which would normally take months to optimize and secure. In addition, these compiled FHE programs automatically scale in performance to the number of cores on a machine without work from the developer. If you want a faster FHE program, just add more cores! No other work is necessary.

The developer watching this talk will learn how to use the compiler to convert C programs into FHE programs, the types of operations they can perform, and performance of the developed programs. In addition, we will show off applications such as secure auctions and secure voting where only the winner of these applications is known. We will demonstrate these applications end to end, enabling developers to see how to take their FHE program idea from concept to actualization.