Open Source Summit Europe 2025

In the last year Zephyr’s cryptographic support saw a lot of changes, the main ones being the deprecation of TinyCrypt and an increased usage of PSA Crypto API following the long standing goal to move toward a common interface for crypto operations.

The main reason for TinyCrypt deprecation is basically that it’s no longer maintained so it cannot get any security fix/improvement nor be adapted to the upcoming PSA interface. This basically leaves Mbed TLS as the only crypto provider in Zephyr, which can either used directly (i.e. building Mbed TLS as library to be linked against other Zephyr’s ones) or being integrated in TF-M image.

This talk will start from an overview of the recent history of Zephyr’s crypto support, presenting reasons for TinyCrypt deprecation and the adoption of PSA Crypto API. The focus will then shift toward Mbed TLS as the only crypto library available today in Zephyr, its layers (TLS and crypto), how it relates to TF-M and the recent improvements added to reduce footprint problems. In the last part of the talk we’ll focus on the future plans for Mbed TLS: what to expect from the next release which is expected in September 2025 and vendor driver support.

Do you find yourself twiddling your thumbs while waiting for dozens of devices to upgrade over Bluetooth? Or do you think twice before pressing the “Firmware Rollout” button due to the anticipated LTE data bill? Introducing Constrained Patch (CPatch), a binary patching algorithm designed specifically for Firmware Over-the-Air (FOTA) upgrades of constrained devices running Zephyr.

In this session I will step through the design and implementation of the algorithm, including:
* Prerequisites for a diff based solution
* Constraints the algorithm works within
* Embedded patching implementation
* Python patch generation
* Comparisons against existing algorithms
* Live demo!

Over the past couple years, the Zephyr documentation has been steadily improving—or so I hope! One thing that has clearly helped is having a yearly check-in where we can openly discuss what’s working, what’s not, and where our priorities should be for the year ahead.

Join this Birds of a Feather session to catch up on some of the recent improvements to the documentation that you may have missed, and bring your suggestions for what we should be focusing on next.

The Cyber Resiliance Act (CRA) will be coming into effect in a few short years. Product makers will have different obligations than open source stewards for compliance. The Zephyr project has been working towards making it easier for product makers to comply with the CRA over the last few years, and will continue to work with the community to refine these capabilities.

From automatic "Build SBOM" generation to LTS Vulnerability fixes being extended to 5 years, the project has some useful starting points.

This talk will discuss what is available, and where some of the gaps will be for product makers and other ecosystem vendors to consider in their roadmaps and support plans.

Switching from any RTOS to Zephyr represents a significant transition. Join us as we revisit our journey that led to the selection of Zephyr and explore our efforts to cultivate a vibrant Zephyr developer community within a large multinational corporation.


In this session, we will discuss:

• Strategies for creating a persuading case for Zephyr

• Techniques for motivating development teams to adopt Zephyr

• Initiatives that contributed to the creation of a thriving developer community

• Ways in which the Zephyr community's momentum can energize your projects

• Methods for building a pool of Zephyr talent for recruitment purposes


Attendees will gain insight into the essential steps required to successfully integrate Zephyr RTOS into a corporate environment and sustain its use over the long term.

Modern Machine Learning makes it possible to automatically extract valuable information from sensor data, and it has become feasible to deploy ML systems to low-cost embedded devices and sensors. This niche is often referred to as "TinyML", and is enabling a range of new applications in consumer electronics, science and industry.

emlearn is an open-source project for deploying Machine Learning models to any device with a C99 compiler. It provides a Python library for converting models made with scikit-learn or Keras to efficient C code. The library has been used for many applications across a range of sensor modes, such as audio, vibration, power-line, radar, et.c.

Zephyr RTOS is a comprehensive open-source operating system that runs on a wide range of microcontrollers. The support for low-power operation, communication protocols, and standardized "sensors" API makes it a very attractive platform for TinyML applications.

In this presentation, we introduce the emlearn project, and show how it can be used together with Zephyr. We will cover the key features and tools that the library provides, and demonstrate how to perform practical Machine Learning tasks.

In the fast-paced realm of embedded systems, rapid product development is essential for market success. Being quick by not having to rewrite code for solved problems was therefore the most interesting promise of Zephyr RTOS when we used for a product prototype of an optical instrument. We'd now like to share what we learned in this context, how exactly Zephyr helped here - especially in the prototyping phase - as well as where it doesn't.

Emphasizing the need for a reassessment of development processes, we’ll discuss the importance of engaging with the Zephyr community and actively participating in the project. Training developers on Zephyr’s functionalities is crucial for maximizing its potential.

The goal is to provide insights into using Zephyr RTOS for efficient product development and encourage well-prepared initiatives, by showing how Zephyr lets one accelerate development while maintaining quality early on.