Help Net Security: Enveil enables organizations to securely train machine learning models

Excerpt:

By allowing customers to leverage Trusted Execution Environments (TEEs) for secure and private model training and evaluation — in addition to existing homomorphic encryption and secure multiparty compute-based offerings — Enveil expands its Secure AI capabilities and solidifies the company’s position as the provider of enterprise-ready solutions powered by Privacy Enhancing Technologies.

The new TEE-powered capability extends the boundary of trusted compute by enabling encrypted federated learning and the secure usage of disparate, decentralized datasets. Organizations can use TEEs, also known as Confidential Computing, to securely train and leverage AI/ML models while ensuring the model development process, the model itself, and the interests of all parties involved remain protected.

“The Secure AI landscape is constantly evolving — now more than ever, there is no such thing as a ‘one-size-fits-all’ solution,” said Dr. Ellison Anne Williams, CEO of Enveil. “We are proud to provide our customers with a range of Privacy Enhancing Technology-powered solutions that address their data-centric AI security challenges, and can adapt within a quickly evolving marketplace. This TEE-enabled expansion is further affirmation of that foundational philosophy.”

As a transformative family of technologies that protect data while it’s being used or processed (Data in Use), Privacy Enhancing Technologies (PETs) uniquely enable users to securely and privately unlock value from data. The core pillars of the PETs category — homomorphic encryption, secure multiparty computation, and trusted execution environments — each have different strengths, which is why Enveil’s core focus is on enabling Secure AI in a manner that successfully addresses customer pain points using the most effective combination of PETs.

“In a new category such as Privacy Enhancing Technologies, there is a tendency to pit technologies against each other, but choosing the right underlying technology depends entirely on the use case requirements, infrastructure, and the desired level and type of protection,” said Dr. Ryan Carr, CTO and VP of Engineering at Enveil. “Organizations looking to utilize PETs should identify their unique requirements to determine the best fit — we are proud our solution suite can deliver a full range of business-enabling options.”

TEEs are hardware-based, secure enclaves that provide data users with a trusted environment where data can be securely processed, protecting the usage of data. In addition to the new TEE offering, users can opt to leverage the homomorphic encryption (HE) and secure multiparty computation (SMPC) capabilities of Enveil’s ZeroReveal ML software solution.

Read full article here.