Niobium brings fully encrypted AI workloads to the cloud with The Fog

A startup called Niobium Microsystems Inc. brings “The Fog” to the cloud, allowing organizations to run artificial intelligence and data processing operations on their most sensitive data without needing to remove it.

While the new platform may sound like 1980s-era horror, Niobium says The Fog is actually something developers will want to embrace. Scheduled to launch later this year, it is the first platform of its kind to use fully homomorphic encryption or FHE. Unlike traditional encryption methods that can only protect data in transit or at rest, FHE makes it possible to encrypt data as it is processed within a server or virtual machine, ensuring that even the most sensitive information remains private at all times.

Niobium claims that Fog offers the benefit of “zero trust”, because the decryption keys will always belong to the owner of the data, ensuring that no third parties, not even itself, can see the information being processed.

The initiative aims to solve a major challenge for organizations, which are sitting on large volumes of sensitive data that cannot be easily accessed for insights. In order to analyze sensitive data, companies often have to decrypt it first, but that means exposing it to potential security breaches and insider threats, as well as violations of compliance regulations. It’s a particular challenge for highly regulated industries like healthcare and financial services, which make them reluctant to move their most sensitive workloads to cloud environments.

CEO Kevin Yoder said many companies simply accept the risk of data exposure as a cost of doing business in the cloud. “The fog ends that trade,” he promised. “Our goal is to make encrypted computing work, grow and be accessible to the groups that need it most.”

Democratizing FHE

Although FHE is not a new invention, the process has traditionally involved large overheads making it slow and expensive to implement at scale. Niobium’s main value proposition is that it overcomes this obstacle. It has developed a new processor called “mystic Core,” based on a programmable gate array design. Available in private beta now, the chip enables FHA functions to run twice as fast as any graphics processing unit or dedicated accelerator.

“Live data encryption has always been very expensive and difficult, limiting applications,” said Holger Mueller of Constellation Research. “For a long time, FHE has been the Achilles heel of data encryption because of this, so it is good to see an innovation with the appropriate name The Fog. The next question is how quickly businesses will adopt this technique.”

To help businesses get started with FHE, Niobium has created a number of template applications. This includes an Encrypted Semantic Search application that makes it possible to query sensitive data for meaning instead of exact matches. It ensures both the query and the underlying data remain fully encrypted, making it ideal for secure loading of production workloads.

There is also a Federated Learning application that makes it possible to train AI models on distributed datasets without exposing them. Finally, the company has developed a Machine Learning Classification application that can analyze encrypted information to identify patterns, trends and security threats.

The launch of The Fog reflects Niobium’s ambitions to transform FHE from a statistical curiosity into an industry-standard model that will pave the way for organizations to do more with their most sensitive data. To that end, the company has launched a dedicated integration suite and software development kit to help developers with no coding experience build apps that can run on the platform.

Fog’s abilities are of particular interest to AI developers. AI systems are only as good as the data that powers them, but large volumes of rich data sets remain disconnected from AI models due to security concerns. Niobium opens up opportunities for companies to train shared AI models on large amounts of aggregated, highly sensitive data, because it can ensure that no party sees the original data sets. It could also solve the immediate privacy question, where employees sometimes upload sensitive data alongside their information so that a large language model can analyze it and generate feedback.

In the future, Niobium says, it will make FHE even more powerful. Today it revealed that it is working with custom silicon developer Semifive US Inc. and Samsung Foundry to create a new integrated circuit specific to the FHE application. First of all this will greatly improve FHE performance, due to the way ASICs circuits can be optimized to run individual applications.

Photos: Niobium