Traditional embedded development is dead. The concept of isolated islands of compute are no longer relevant and instead all applications should be conceived as “edge processing” with a tether, thick or thin, back to a centralized processing resource for analysis. Or at least that is the theory propagated across the industry in recent months. With applications rapidly transitioning from traditional embedded devices to cloud orientated edge compute nodes, this article looks at how the approach must change there to address the growing need to integrate cloud connectivity into both the development and manufacturing processes, and to create a seamless secure supply chain for the internet of things (IoT).
The reality is of course far more nuanced and fragmented, with the embedded marketplace taking many years to integrate new concepts and having little appetite to sign up for decades of committed support for flash-in-the-pan technologies. However, there is a clear reality that many of our embedded systems are making the transition from disconnected islands through to edge compute, and that this is driving performance requirements, communication stack integration, artificial intelligence / machine learning (AI/ML), and many other features on next generation applications.
This connectivity transition is pushing security requirements and the compliance frameworks required for consumer, industrial and critical national infrastructure marketplaces, as every connection is also an attack vector into the system.
There are many articles dedicated to the benefits and challenges associated with edge orientated compute, where resources should sit, where data should be pre-processed, and to what extent. The balance between determinacy and lag, system availability and robustness to outages is also a critical decision, especially following the impacts of the major social media outages in October 2021, where many connected devices failed to operate successfully once DNS flaws were exposed. However, many of these topics are application specific, and whilst better tools can help, ultimately these are design decisions driven by cost, time to market, and complexity.
From a developer perspective, cloud connectivity is a major headache, partly because security and connectivity integrate many different concepts, and partly because the requirements of the customers are evolving. The simple challenge of “cloud connectivity” rapidly breaks down into the need for a secure communications stack, root of trust and identity challenges, certificate injection, secure boot and update mechanisms, and a myriad of other demands.
However, this is just the developer side and other challenges of similar size remain for production, device services, and the enterprise. For the enterprise, new legislative challenges around lifecycle support, vulnerability disclosure and update management deliver business model challenges, with no more “fire and forget” sales and shipping. For device services the need to ultimately integrate into a customers’ security operation center (SOC), with the requirement to deliver formal authentication and measurable security across confidentiality, integrity, and availability, is critical.
Production is a particularly challenging aspect of cloud and security integration, impacting systems geared to homogeneous and repeatable manufacture, with the need for truly unique cryptographic identities ensuring each is universally unique, whilst still having standard certificate structures, based for example on x.509 certificates. This challenge is further exacerbated by ensuring that these identities are cloud-ready, and that devices are capable of being onboarded into whichever cloud compute network the user desires.
This onboarding has been highlighted by many cloud vendors as being the single biggest pain point in their flow, both to them and to customers, and the biggest challenge to CISO’s (chief information & security officer) integrating IoT devices into internal networks. In fact, provisioning, to ensure a sufficient identity, ownership and defined cybersecurity risk, is the biggest single inhibitor to the rise of IoT in industrial, transportation, and infrastructure applications.
To address this production issue, of homogeneous manufacturing with differentiate identity and root of trust, there are two major alternatives.
The first is to integrate a separate SIM (subscriber identity module), like that seen in mobile phones. However, this does require the integration of a SIM card, with size and logistical issues, or alternatively the implementation of an iSIM (integrated SIM) which requires an additional processor built into the silicon, with specific cost implications.
The alternative is to simply lock down the device, with a robust secure boot manager (SBM), to ensure that all code operating on the device has been produced and injected securely, with robust code authentication. The advantage of this is low-cost impact, with a small amount of memory used to extend the boot process, and a far wider availability of devices, including many popular devices used today, ensuring rapid engineering adoption and a simple release cycle. The flexibility of the solution supports self-signing certificate structures for organizations that wish to own the device identity, or third-party certificate frameworks for those looking to outsource.
The solution demonstrated today leverages the SBM approach, ensuring that a huge array of devices can be simply engineered for security, and a standard certificate for Azure cloud connectivity injected. This certificate framework enables developers to rapidly design for millions of devices to be produced with unique identities, and then ensures these credentials are simply valeted into the Azure cloud service. This ensures that all devices produced are identified into the cloud, ensuring that the Azure service has knowledge of the OEM vendor, and device capabilities; and ensures that the end user has confidence that the device integrates easily, that the device has been produced correctly, and that the device has been released as part of a secure supply chain, reducing purchasing risk and integration costs.
Fundamentally, as stated, this new cloud orientated flow builds on existing development processes, which are operating with major partners and customers today, and leverages standard industry production mechanisms. This means the solution has practically zero impact to modern production flows, is available via major electronic component distributors, and is accessible across a wide range of devices. There is no upper or lower limit on device numbers, ensuring a simple pathway to market for simple prototypes and early market adoption through to high-volume applications.
This is the approach taken by Secure Thingz, an IAR Systems company, enabling this new capability to enable cloud provisioning and onboarding, in partnership with Microsoft Azure. As described, this enables rapid development, volume manufacturing, and integrated update management. This essentially provides a valuable extension to secure provisioning technology, allowing organizations to easily integrate their devices en-masse onto the cloud, removing traditional onboarding challenges of unknown devices for both customers and cloud services.
In summary, connecting and onboarding devices to the cloud is probably the single biggest challenge to the industry right now, primarily due to the number of partners and device types. By leveraging an extended flow as described it is now possible to design and provision “cloud ready” for Microsoft Azure, reducing development costs, reducing customer integration challenges, and removing cybersecurity risks associated with adding devices to networks.
Haydn Povey is the chief strategy officer at IAR Systems as well as founder & CEO of Secure Thingz. He also sits on the executive steering board of the IoT Security Foundation. With over 20 years in management positions at global technology companies, Povey was also at Arm for 10 years where his roles included strategy and product roadmaps for IoT and M2M security, and leading the development and introduction of the Cortex-M microprocessor family.
- How to secure IoT connectivity and onboarding without writing code
- Qualcomm defines 30 verticals in IoT as-a-service offer
- Why TCP tunneling matters in IoT
- A guide to securely connecting IoT devices to the cloud
- EdgeLock 2GO IoT platform securely connects and manages edge devices
- Facilitating IoT provisioning at scale
- IoT security hinges on effective device enrollment with public key infrastructure