Enveil has released its new encrypted training solution, ZeroReveal ML Encrypted Training (ZMET).
The enterprise-ready product extends the boundaries of trusted computing by enabling encrypted federated learning and the secure use of disparate, decentralized datasets for machine learning applications.
Designed to solve specific customer problems, ZMET enables organizations to train models in an encrypted manner while ensuring that the model development process, the model itself, and the interests of all involved parties remain protected. The product expansion, an extension of Enveil’s suite of machine learning solutions, follows the company’s announcement of $25 million in Series B funding.
The rise of the digital economy is driving a vast market need to bridge global data silos and extract insights through secure and private data usage, analytics, and machine learning. Enveil’s ZeroReveal solutions empower this digital transformation by changing the paradigm of how and where organizations can leverage data to unlock value.
A recent report by Gartner “Innovation Insight for Federated Machine Learning”, which recognizes Enveil as a representative vendor, highlights this market dynamic: “By 2025, 80% of the world’s largest organizations will have participated in machine learning at least once. Federated Automatic (FedML). to create more accurate, safe and environmentally sustainable models” (March 2022).
“Today’s digitally driven business landscape demands solutions that extend an organization’s reach without sacrificing privacy or security,” said Dr. Ellison Anne Williams, Founder and CEO of Enveil. “By ensuring that models are trained securely – and that the model itself and its associated results remain encrypted – ZeroReveal Machine Learning enables organizations to leverage ML to securely obtain insights from sources of data across silos, jurisdictions or borders, even when using highly sensitive models or training data.”
ZMET uses advances in privacy-enhancing technologies, namely Secure Multi-Party Computing (SMPC), to train models in an encrypted manner. This encrypted training process enables secure federated learning, protecting the model development process, the data used for training, and the interests and intent of the parties involved.
Organizations can confidently leverage sensitive data and/or ML models during training without the risk of exposure, providing enhanced models that can be used more accurately to gain insights and deliver value. Models can be trained using data sources across security domains and organizational boundaries without the risk of unintended exposure.
“We are proud to be the first in our category to offer an encrypted training product with a concrete and verifiable security posture: ZMET offers an unparalleled ability to derive insights from data without the need to trust other parties. while computing,” said Dr. Ryan Carr, chief technology officer at Enveil. “These privacy-preserving machine learning training capabilities are grounded in the needs of our customers, designed to overcome barriers and add business and mission value for today’s ML and data science use cases. .”
At its core, ZeroReveal Machine Learning is a bipartite proxy layer software system enabling decentralized, distributed evaluation and training of encrypted machine learning models across multiple datasets. Enveil protects the content of the search, analysis or machine learning model – and its corresponding results.
The company’s decentralized approach enables data to be operated securely between entities and across organizational, jurisdictional and security boundaries, expanding the usefulness of data without the need to move or consolidate sensitive assets.