Machine learning project from DARPA gets nod for Phase 2
NewsJanuary 10, 2023
HOBOKEN, N.J. Data-privacy firm Duality Technologies has been selected for Phase II of the Defense Advanced Research Projects Agency (DARPA) Data Protection in Virtual Environments (DPRIVE) program.
The DARPA project is working on developing capabilities for accelerated privacy-protected machine learning (ML), which will then get integrated into Duality Technologies' secure data-collaboration platform, with the goal of making ML on encrypted text as fast as ML on unencrypted data by designing a hardware accelerator integrated with advanced FHE [fully homomorphic encryption] software such as OpenFHE.
According to information from DARPA, FHE enables computation on encrypted information, enabling users to balance the use of sensitive data to its full extent while minimizing the risk of data exposure; FHE does require a prohibitive amount of compute power and time.
In the initial phase of the DARPA program, a Duality team developed a novel application-specific integrated circuit (ASIC) -- called TREBUCHET -- that accelerated FHE computations. The TREBUCHET team will continue to leverage the Duality-powered OpenFHE, an open source FHE software library that treats computationon encrypted data in a new way, thereby ensuring data security to address applications for the U.S. government, U.S. Department of Defense (DoD), and commercial applications.
