SaTC: CORE: Medium: Towards a Universal Cryptographic Acceleration System

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

Cryptographic technologies for data encryption and authentication are mature and pervasive on the Internet. These cryptographic techniques protect data only during transmission and storage, not processing. Cloud and distributed services process sensitive data in a variety of ways, from medical diagnostics to tax returns to financial transactions. Servers must decrypt data before processing it, and once data is decrypted, it can be stolen by an attacker. Further, the computation performed on the data can be corrupted or maliciously altered, producing incorrect results with catastrophic consequences, e.g., an incorrect cancer diagnosis. This project will develop a performant, cryptography-based approach to secure computation that provides privacy and integrity guarantees, without trusting the cloud infrastructure. Sophisticated cryptographic primitives like Fully Homomorphic Encryption (FHE), Verifiable Computation (VC), and Private Information Retrieval (PIR) are used in applications such as outsourced computation, cryptocurrencies, private search, and anonymous communication. These primitives provide privacy and integrity guarantees, without requiring trust in cloud infrastructure. However, current implementations suffer from poor performance, limiting their use. This project takes a synergistic approach of designing programmable hardware for cryptography and tailoring cryptography to this hardware, leveraging lattice-based cryptography and computation-communication tradeoffs. The overall goal of this project is to accelerate cryptographic schemes for FHE, VC, and PIR using a single hardware accelerator. The proposed approach can accelerate cryptographic primitives by orders of magnitude, enabling important use cases. Moreover, this project will develop compiler techniques that abstract the complexity of these cryptographic primitives, making them easier to use, and automatically target this hardware accelerator. The project will open-source the proposed hardware accelerator and compiler that can be shared across the whole research community. 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.