SaTC: CORE: Small: Compiler Support for Scale Management of CKKS Fully Homomorphic Encryption

Fully Homomorphic Encryption (FHE) allows arithmetic operations on encrypted data without decryption, ensuring that the decrypted result matches the outcome on unencrypted data. FHE has potential to enable privacy-preserving machine learning services in regulated fields like healthcare, finance, and insurance. However, managing ciphertext scales is challenging for programmers. This project proposes new FHE compiler techniques that support automatic scale management to ease programming and optimize latency and accuracy. If successful, these techniques will allow programmers easily and fully exploit recent advances in FHE schemes, significantly improving data privacy during computation offloading. This project includes three research tasks. In Task 1, the team of researchers proposes a new scale management solution to minimize FHE program latency without costly iterative exploration. In Task 2, the team will introduce an automatic bootstrapping management technique to optimize bootstrapping placement and reduce latency. In Task 3, the team plans to develop a new extensible hardware abstraction to handle diverse hardware accelerators and their libraries, improving integration with FHE compilers and accelerators. The results of the project will be available as open source documents and tools. Educational curriculum will be created to educate students and train practitioners. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.