The Defense Advanced Research Projects Agency has awarded cryptography and data sciences company Duality Technologies a $14.5 million contract to help develop an encryption concept for privacy-sensitive applications such as health care and finance.
Duality Technologies said Wednesday it will lead a team effort to create a hardware accelerator that uses Fully Homomorphic Encryption techniques and an Application-specific Integrated Circuit code known as TREBUCHET as part of DARPA's Data Protection in Virtual Environments program.
The team includes Carnegie Mellon University, Drexel University, New York University, TwoSix Labs, SpiralGen and the University of Southern California Information Sciences Institute.
The FHE-based prototype hardware is envisioned to train machine learning algorithms on handling encrypted data including organizations' sensitive or confidential information and intellectual property.
The accelerator will also be integrated with the open-source PALISADE software library of FHE concepts which are commonly used for data privacy purposes.
Kurt Rohloff, cofounder and chief technology officer of Duality Technologies, noted that the award represents the company's achievement of accelerating encrypted neural networks for DARPA.
David Cousins, director of the company's Duality Labs and principal investigator for the effort, added that the contract builds on the company's data encryption work with DARPA over a decade ago.
Duality Technologies offers a suite of cryptography products that support applications such as health care and life sciences, financial operations, insurance and privacy compliance.