DARPA Contracts with Duality Technologies to Develop Privacy-Preserving ML

Duality Technologies partners with Harvard Medical School and Two Six Labs to accelerate privacy-preserving, collaborative machine learning protected through Homomorphic Encryption. This capability supports research into COVID-19 susceptibility while keeping patient data private and secure.

Duality Technologies, a leading provider of Privacy-Enhancing Technologies (PETs), announced that it has contracted with the Defense Advanced Research Projects Agency(DARPA) to develop a privacy-preserving Machine Learning (ML) capability that can train models on encrypted data from multiple sources. The capabilities developed under the contract will be applied to researching genomic susceptibility to severe COVID-19 symptoms in a manner which preserves individual privacy.

The development is based on Homomorphic Encryption (HE), a privacy-enhancing technology that allows multiple parties to analyze encrypted data and gain insights without exposing Personally Identifiable Information (PII). Duality, at the forefront of industrial research and practical implementation of HE, is leveraging its capabilities in partnership with Harvard Medical School and Two Six Labs, a research and development company in the field of data science and cybersecurity, to support Genome-Wide Association Studies (GWAS) related to COVID-19 research.

Recommended AI News: Fetch Robotics And Piedmont National Partner With Puro Lighting

Duality researchers have already established that applying Homomorphic Encryption to large-scale GWAS can yield results 30 times faster than alternative approaches based on Secure Multiparty Computation. GWAS assist medical research teams in understanding whether the presence of specific genomic variants in an infected person’s DNA affects their likelihood of developing severe COVID-19 symptoms. These insights may significantly improve healthcare providers’ ability to treat COVID-19 patients, yet are currently difficult to obtain as they require large sets of personal data on which predictive models can be trained. With the new machine learning capability that runs on encrypted data, researchers can pool sensitive clinical and genomic data from a variety of healthcare institutions to build predictive models while complying with global data protection and privacy regulations.

“This contract award supports Duality’s commitment to promoting responsible, privacy-preserving use of big data for the public good and for national security,” said Dr. Kurt Rohloff, Co-founder and CTO, Duality Technologies. “Secure Machine Learning model training is the next frontier in privacy-preserving analytics, and we are excited to make this advanced capability applicable to real-world challenges.”

The DARPA contract will also support research into cyber threat intelligence sharing – also based on Homomorphic Encryption. At present, most research into cybersecurity threats and breaches in the private sector is conducted by competing companies, which are often hesitant to reveal sensitive details about cyber-attacks on company networks. This diminishes their ability to develop threat detection models and to strengthen collective and individual resilience. By securely sharing and generating insights from each other’s data while it remains encrypted, multiple companies and agencies can build smarter detection mechanisms without exposing private information – enabling a paradigm shift in the global fight against ever more sophisticated cybercrime.

Recommended AI News: IPsoft’s Amelia Now Available On Genesys AppFoundry

DARPA has previously supported research and development in HE, sponsoring the establishment of PALISADE, the open-source HE library co-founded by Dr. Kurt Rohloff with Dr. Yuriy Polyakov and Dr. Dave Cousins, both Principal Scientists at Duality Technologies. Duality, together with Intel, Microsoft, Samsung SDS, and IBM, is also a co-founder ofthe international Homomorphic Encryption Consortium that brings together academia, government and industry leaders to advance industry standards for HE, accelerating market adoption of privacy-preserving computing technologies.