Duality Secures $14M DARPA Contract for Homomorphic Encryption

The Challenge of AI Training and Data Security

The training of artificial intelligence systems is now a cornerstone of the modern technology landscape. However, the sheer volume of data required for effective AI training presents significant security risks.

The Defense Advanced Research Projects Agency (DARPA) is actively seeking solutions to mitigate these risks. They have awarded a $14.5 million contract to Duality Technologies to pioneer a new approach.

Duality's Expertise in Homomorphic Encryption

Duality specializes in fully homomorphic encryption (FHE), a sophisticated encryption method. Traditional encryption renders data unreadable without the decryption key.

While this protects data, decryption is computationally demanding, especially with large datasets. Furthermore, decrypted data becomes vulnerable to breaches and misuse.

How FHE Offers a Solution

Fully homomorphic encryption allows for data analysis and manipulation while the data remains encrypted. This eliminates the need for decryption and the associated security risks.

However, FHE is currently more computationally intensive than standard encryption. This limits its practicality for applications involving substantial data volumes.

DARPA's DPRIVE Program and the TREBUCHET Chip

DARPA’s interest in this field is substantial, backed by significant funding. This contract falls under the DPRIVE (Data Protection in Virtual Environments) program.

The primary objective is to develop a dedicated chip – an ASIC codenamed TREBUCHET – to dramatically accelerate FHE processing. The goal is to achieve a performance increase of at least tenfold.

Collaboration with Leading Institutions

Duality will collaborate with experts from several prominent universities and research labs, including:

• University of Southern California (USC)
• New York University (NYU)
• Carnegie Mellon University (CMU)
• SpiralGen
• Drexel University
• TwoSix Labs

Duality has a proven track record in this area, having previously collaborated with DARPA and contributed to the development of early HE hardware accelerators.

“Our team members have been supporting DARPA-funded innovation and application of FHE for over a decade,” stated David Bruce Cousins, Duality Labs director and principal investigator. “We developed the first HE hardware accelerators under the DARPA PROCEED program in 2010 and are lead developers for the PALISADE open source FHE library, initially created for the DARPA SAFEWARE program in 2015.”

Looking Ahead

The project’s timeline remains uncertain. Given the complexities of the technology, tangible results are not anticipated for at least two to three years.

This initiative represents a crucial step towards securing sensitive data while enabling the continued advancement of artificial intelligence.