Vaultree Open-sourced its Data Encryption Technology Elements

Highlights:

• According to a Deloitte LLP survey earlier this year, data privacy was ranked as one of the top three concerns concerning generative AI by 72% of IT professionals, up from 22% the previous year.
• An internal FHE library, Vaultree Encrypted Numerical Python, that facilitates scalable and secure ML operations, is a segment of Vaultree’s VENum technology.

Data encryption startup company Vaultree Ltd. open-sourced its data encryption technology stack elements to facilitate operations on encrypted data, eliminating the need for decryption.

The solution developed by the Cork, Ireland-based company is intended to overcome the scalability issues of Fully Homomorphic Encryption schemes, which allow for arbitrary calculations on encrypted data, including infinite adds and multiplications.

By using an encryption strategy that transforms plaintext and other data types into a ciphertext that can only be decrypted with a decryption key, homomorphic encryption operates. It makes it possible for calculations to match the results of plaintext operations without any need for a decryption step. This makes it helpful in situations when protecting privacy is crucial, including those involving financial or medical details.

FHE is slower and requires more computing power than other homomorphic encryption methods, but it is more flexible.

With the training of generative AI models, the privacy issue is becoming more considerable.
According to a Deloitte LLP survey earlier this year, data privacy was ranked as one of the top three concerns regarding generative AI by 72% of IT professionals, up from 22% the previous year.

Nurturing Feedback

According to the Co-founder and CEO Ryan Lasmaili, the firm made the decision to make its inventions—for which it has already been granted 27 patents—open-source in order to promote openness and solicit input from the encryption community.

“The largest tech companies have been trying to solve the FHE scalability problem for 40 years and a lot of claims have been made,” he said. “We’re interested in being transparent about what works and how it solves a problem.”

In a study published this summer, Vaultree explained its methodology. Irrespective of the dataset’s size, its approaching methodology builds a single key structure to enable consistent ciphertext size and execution period for encryptions. The method is just 10% to 15% slower than plaintext operations and tackles the size and noise accumulation problems that have historically impeded scalability and multiuser setups. According to Lasmaili, that’s a big improvement over some FHE techniques, which can be up to 40 times slower.

Earlier approaches to FHE “have been developed by brilliant minds but only in an academic setting, and only a handful of people know these technologies very well,” he said. “You have to go back to the drawing board and solve the math problems of creating scalable and production-ready FHE.”

Python Libraries

An internal FHE library, Vaultree Encrypted Numerical Python, that facilitates scalable and secure ML operations, is a segment of Vaultree’s VENum technology. A well-known open-source Python toolkit for numerical calculation, Numpy provides a large number of mathematical operations in addition to multidimensional arrays and matrices.

Another component is VENum Machine Learning, a Python library that focuses on machine learning and makes it possible for users without sophisticated data science knowledge to safely complete complex ML tasks. It is built on Vaultree’s encryption technique. “We don’t expect cryptographers to get data science degrees,” Lasmaili said.

For improved information discovery, VENum enables encrypted file searching and ranking. Financial institutions can safely share encrypted data for use cases like fraud protection, while healthcare institutions can use it to pool patient data for modeling without compromising privacy.

According to the corporation, their technology is made to have as little of an influence on performance as possible. The library supports tabular data, images, graphs, unstructured data, and time-series. It connects to renowned key management platforms such as HashiCorp Inc.’s Vault and Google LLC’s Cloud Key Management Service. This guarantees that data stays protected while enabling dynamic management of encryption keys.

Vaultree, which has garnered USD 16 million, uses its now-open-sourced technology to sell exclusive tools and services. According to Lasmaili, the company would allow community input to guide its growth plan. Support for vector databases and multiplication depth—the quantity of sequential multiplication operations carried out throughout a computational process—are features of the current roadmap.

Lasmaili stated enabling encrypted data in model training resolves another restraints. “Large language models will exhaust available public data by 2026,” he said. “Now we can use private data.”