Announcements

Encrypted computing from edge to cloud with NXP

12/11/26
·
Marc Desgroseilliers

Connected devices generate growing volumes of data, much of which is sent to the cloud for analysis. Over the past few months, our team has been working with the cryptography experts at NXP® Semiconductors to explore how this data can remain encrypted during cloud processing, helping to reduce exposure throughout its journey from device to cloud and back.

The gap between the device and the cloud

Modern vehicles and industrial machines continuously generate data. Analysing that data in the cloud can help predict component failures, detect faults early and plan maintenance across an entire fleet. Yet it can also reveal sensitive information, from proprietary manufacturing processes to commercially confidential operational data.

Today, encryption typically protects data in transit and at rest. In conventional cloud processing, however, the server must decrypt the data before analysing it. This creates a potential point of exposure to attackers, malicious insiders or a cloud provider with whom the data owner does not wish to share the underlying information. As more devices become connected and more operations rely on cloud-based analysis, protecting data during processing becomes increasingly important.

Fully homomorphic encryption (FHE) helps address this gap by enabling computations, including machine learning inference, to run directly on encrypted data. The server processes encrypted inputs and returns encrypted results without needing the decryption key. This allows a data owner to use cloud computing resources while keeping the underlying data confidential.

Two ends of the same problem

Encrypted computing involves a client and a server, each with a distinct role. The client, often a device with limited memory and power, encrypts the data and manages the keys. The server performs computations on the encrypted data, handling most of the computational workload.

NXP has deep expertise on the client side. Its chips are used across the automotive, industrial and IoT, mobile and communications infrastructure markets, and many of them include dedicated features to protect keys and data on the device.

Belfort focuses primarily on server-side processing. Computing on encrypted data is more computationally demanding than processing unencrypted data. Our goal is to use GPUs to make these workloads faster and more cost-effective. Our eML platform brings this approach to machine learning, enabling models to process encrypted inputs and produce encrypted results without requiring the server to decrypt the data.

What we are exploring together

Together, we are investigating how encrypted computing could help protect data from connected devices, including vehicles and industrial systems, during cloud-based analysis. One scenario under consideration is for a device built on NXP technology to encrypt data locally, Belfort’s eML platform to perform machine learning inference on the encrypted inputs, and the encrypted results to be returned for decryption by the data owner.

The work is at an early stage. Processing encrypted data is more computationally demanding than processing unencrypted data, so a key objective is to understand which workloads could be practical today and where further performance improvements are needed. Combining NXP’s expertise in applied cryptography and embedded security with Belfort’s focus on encrypted processing helps us assess both device-side requirements and cloud-side performance.

“Protecting sensitive data is important throughout its journey, including when it is being processed. Through our work with Belfort, we are exploring how encrypted computing could complement device security and help protect data beyond the edge. Our focus is on understanding the practical requirements and performance trade-offs for automotive and industrial applications.”

Joppe Bos, Technical Director and Cryptographer, NXP Semiconductors

What's next

We will share further updates as the work progresses. Organisations that work with sensitive data from connected devices and want to explore what encrypted computing can offer are welcome to start building on eML at eml.belfortlabs.com or to contact us at belfortlabs.com/contact.

‍

About Belfort

Belfort is building the foundation for trustless intelligence. Belfort builds encrypted computing (FHE) infrastructure so sensitive data can be processed without ever being decrypted. Spun out of KU Leuven’s COSIC lab, its hardware-software co-design makes this practical at scale. The company has offices in San Francisco, USA, and Leuven, Belgium. https://belfortlabs.com/

Media contact: Laurens De Poorter, laurens@belfortlabs.com

FAQ