New Jersey-based Duality Technologies and Intel are collaborating to offer accelerated homomorphic encryption-based machine-learning applications on Amazon Web Services (AWS).

This will bring privacy-enhancing capabilities, available at scale on large data sets, supporting a variety of machine-learning (ML) workloads involving sensitive data in the cloud.

Duality provides privacy-enhanced collaborative data science and ML based on homomorphic encryption (HE). Its privacy-enhanced SecurePlus ML applications are now available to AWS customers fuelled by third-generation Intel Xeon processors, using accelerated HE.

These applications allow AWS customers to process new ML workloads in the cloud at scale while protecting sensitive data and preserving privacy. Intel’s latest Xeon scalable processor includes the HE acceleration layer HeXL, which uses Intel vector extensions that accelerate AI workloads and come equipped with hardware-enhanced security capabilities. This is the first time that such capabilities are available at scale for large data sets and for various ML models, seen as a major step for the adoption of privacy-enhancing technologies.

The collaboration between Duality and Intel enables accelerated privacy-enhanced ML workloads on AWS, allowing enterprises to apply data science and ML computations on their sensitive data in the cloud, while the data remain encrypted at all times, even during analysis. This capability also ensures that organisations processing sensitive data in the cloud remain compliant with data privacy regulations as they are never directly exposed to the data being processed.

Earlier this year, Duality and Intel teamed up to deliver optimised data science and AI applications for privacy-enhanced data collaboration. The collaboration focused on accelerating ML applications with HE on the new third-generation Xeon scalable processors.

“Duality and Intel are working together to allow organisations across the globe to benefit from secure, AI-powered data collaboration at scale, enabling a whole new world of big data insights and privacy-protected collaboration across industries,” said Alon Kaufman, CEO of Duality Technologies. “We are excited to bolster our relationship with Intel and allow AWS customers to use the cloud for ML workloads on sensitive data, removing a long-standing roadblock to utilising the cloud within regulated industries.”

Founded by cryptographers and data scientists, Duality harnessed its foremost expertise in HE and data science to develop the SecurePlus suite of secure data collaboration applications, addressing the needs of organisations that seek to collaborate on sensitive data while protecting their business interests and complying with data privacy regulations. SecurePlus applications enable such collaboration on protected data and AI models by keeping them encrypted throughout computation. The collaborating parties can run computations securely in any environment – whether their own, their partners’ or on a third-party cloud – without exposing sensitive data, such as personally identifiable information (PII).

The Intel HE toolkit, recently released as an open-source project and available at the Intel Developer Zone, is the prime vehicle for the continuous delivery of this Intel HE technological innovation to users. The toolkit has been designed with usability in mind to make it easier to evaluate and deploy HE technology on Intel platforms.

“At Intel, we believe that security and privacy are critical to today’s data-driven economy,” said Nir Peled, general manager of private AI and analytics at Intel. “Privacy-enhancing homomorphic encryption, now available at scale on AWS, will fundamentally change how data are shared and leveraged by organisations worldwide. We look forward to continuing working with Duality to make privacy, data confidentiality and data utility achievable goals for all enterprises.”