Intel has joined the Defense Advanced Research Projects Agency (DARPA) Data Protection in Virtual Environments (DPRIVE) program. The chip company will work with Microsoft to drive fully homomorphic encryption (FHE) development.
As a DPRIVE contributor, Intel will design an application-specific integrated circuit (ASIC) accelerator to reduce the performance overhead associated with FHE, the company said. The accelerator could improve FHE workload execution over existing CPU-driven systems and reduce cryptograms' processing time by five orders of magnitude.
In addition, Intel will assess its progress against pre-established performance targets on AI training and inference workloads using homomorphically encrypted data during the program, the company stated. It also will work with Microsoft and international standards bodies to develop FHE guidelines.
What Is DPRIVE?
DPRIVE is a multiyear program that focuses on finding solutions to provide computing on fully encrypted data without access to decryption keys, Intel said. Along with Intel's research team, DPRIVE is backed by research teams from Duality Technologies, Galois and SRI International.
Research teams will develop an FHE accelerator hardware and software stack that reduces the computational overhead required to make FHE calculations, DARPA noted. They also will create accelerator architectures that are flexible, scalable and programmable and explore various approaches with different native word sizes.
Furthermore, research teams are exploring new approaches to memory management, data structures and programming models and verification methods, DARPA said. This ensures that research teams can validate that FHE implementation is correct-by-design.
A Closer Look at FHE
FHE is a data security approach that uses cryptography to provide mathematical proof of encryption, according to DARPA. In doing so, FHE eliminates the need for decryption.
Traditional encryption protects data while stored or in transmission but requires decryption to perform a computation, analyze it or employ it to train a machine learning model, DARPA pointed out. Meanwhile, decryption increases the risk of a data compromise or leak.
Comparatively, FHE enables computation on encrypted information, DARPA indicated. It empowers users to get the most value out of their sensitive data while minimizing its exposure to security risks.
IBM Releases FHE Offerings
FHE research is ongoing, and IBM Security has already integrated several FHE offerings into its portfolio.
IBM in December 2o20 launched services that allow companies to experiment with FHE. It also unveiled an FHE toolkit designed to help organizations protect data on Apple macOS and iOS operating systems.