Hardware startup product security: a worked example

Hephaestus Technology8 min read

Hardware startup product security at seed stage means making a small number of decisions early, while they are still cheap: choosing a chip with secure boot, giving every device unique credentials, signing firmware updates, generating an SBOM, publishing a disclosure policy, committing to a support period and setting up a 24-hour reporting route. This article walks through those decisions for a clearly hypothetical startup building a Wi-Fi sensor for the EU and UK markets. The rules behind them are the Radio Equipment Directive's cybersecurity requirements today and the Cyber Resilience Act from December 2027.

Key takeaways

  • Map the rules first: RED cybersecurity applies to wireless devices now, and the CRA takes over on 11 December 2027.
  • Hardware choices made before the first PCB run, such as secure boot and a secure element, are the hardest to change later.
  • Unique per-device credentials and signed OTA updates satisfy several EU and UK requirements at once.
  • A support period of at least five years is a business commitment that affects component choice, cloud costs and pricing.
  • Set up CRA reporting before the first unit ships, because it already applies to products on the market.

Who is the hypothetical startup?

The company in this example is fictional. Call it Brightwell Sensors, an invented name: a six-person seed-stage startup in Dublin building an indoor air-quality sensor for offices and homes. The device connects over Wi-Fi to a cloud service, and users view readings in a mobile app. Brightwell plans a pilot of a few hundred units in early 2027 and a wider EU and UK launch later that year. It has one firmware engineer, one cloud and app developer, and a contract hardware designer.

Nothing in this example describes a real company, client or result. It is a way to show the order in which security decisions arise.

Step 1: which rules apply, and when?

Brightwell starts with a one-page regulatory map, because the dates drive the plan.

Rule Applies to Brightwell because Key date
RED cybersecurity, Delegated Regulation (EU) 2022/30 Internet-connected radio equipment processing personal data Applies now (since 1 August 2025)
CRA reporting, Article 14 Any product with digital elements on the EU market, including pilot units sold commercially Applies now (since 11 September 2026)
CRA full requirements Units placed on the market from that date 11 December 2027
RED cybersecurity repeal, Delegated Regulation (EU) 2026/339 CRA replaces RED cybersecurity rules 11 December 2027
UK product security regime (GOV.UK) Consumer connectable product sold in Great Britain Applies now (since 29 April 2024)

The conclusion is simple: Brightwell will launch under RED and keep shipping under the CRA, so it designs to the stricter CRA requirements from the start and documents RED compliance for launch. Our news analysis of CE marking for connected products explains the handover between the two regimes.

It also checks classification. An air-quality sensor is not listed in Annex III as an important product, unlike smart home security products such as cameras and door locks, so it is a default product and can use internal control for CRA conformity assessment. If Brightwell later adds an alarm function, that assessment must be repeated.

Step 2: what does the risk assessment cover?

The CRA requires a documented cybersecurity risk assessment that is updated through the support period (Article 13(3)). Brightwell's first version fits on a few pages and asks four questions:

  1. What could an attacker want? Access to the home or office network, the user's account, location or occupancy patterns inferred from air data, or a botnet node.
  2. Where could they get in? Wi-Fi provisioning, the local debug port, the cloud API, the mobile app, the update channel and the supply chain of firmware components.
  3. What is the impact? Privacy harm to users, damage to other devices on the network, and loss of trust.
  4. Which Annex I requirements does each risk map to? For example, secure by default, access control, encryption, attack-surface reduction and secure updates.

The output is a short list of controls, which become the next steps.

Step 3: which hardware decisions must happen before the first PCB?

Three, because they are expensive to change after tooling:

  • A microcontroller with secure boot and protected key storage. ETSI EN 303 645 provision 5.7-1 recommends verifying software with secure boot, with a hardware root of trust as one way to provide it (ETSI).
  • Either a secure element or on-chip protected storage for keys. Provision 5.4-1 requires sensitive security parameters in persistent storage to be stored securely.
  • A plan to lock debug interfaces in production units, so JTAG or UART cannot bypass the rest.

Brightwell's hardware designer adds a production step to burn fuses and disable debug at the end-of-line test.

Step 4: how are devices given credentials?

Each unit gets unique credentials at manufacture, and there is no universal default password. This one decision satisfies several rules: ETSI provision 5.1 ("no universal default passwords"), the UK regime's ban on universal default and easily guessable passwords (GOV.UK), and the EN 18031 citation, which gives no presumption of conformity where "the user is allowed not to set and use any password" (Implementing Decision (EU) 2025/138).

Brightwell provisions a per-device certificate during factory test and uses it to authenticate the device to the cloud. Users pair through the app; there is no local web interface with a login.

Step 5: how does firmware get updated safely?

With signed images, verification before installation, anti-rollback and A/B slots. The CRA requires security updates, automatic by default where applicable, and secure distribution (Annex I). Brightwell keeps its signing key in a cloud HSM that only two people can use, never on the CI server. Our secure OTA firmware update guide gives the full step list Brightwell follows.

Step 6: what goes into the build pipeline?

An SBOM for every firmware and app release, and an automated vulnerability scan against it. The CRA requires an SBOM in a commonly used, machine-readable format covering at least top-level dependencies (Annex I Part II). The 2026 Minimum Elements published by CISA and partners add component hashes and licences to the recommended fields (CISA). Brightwell lists its RTOS, Wi-Fi SDK and TLS library by hand in a manifest, because the package manager cannot see them.

Step 7: how will Brightwell hear about vulnerabilities?

Through a published disclosure policy, a security contact and a security.txt file. The CRA requires a coordinated vulnerability disclosure policy and a contact address, and ETSI provision 5.2-1 asks for acknowledgement and status-update timelines. Brightwell adapts our vulnerability disclosure policy template, routes the security inbox to the two founders, and commits to acknowledging reports within three business days.

Step 8: how long will Brightwell support the product?

Five years from the date each unit is placed on the market, published on the box and website. Article 13(8) requires a support period of at least five years unless the product is expected to be in use for less, and the UK regime requires the minimum security update period to be published.

This is the decision with the biggest business impact. Five years of updates means five years of cloud hosting, certificate renewal, a maintained build environment and someone to triage reports. Brightwell builds those costs into its pricing model and investor plan now, rather than discovering them in year three.

Step 9: what happens if a vulnerability is exploited?

Brightwell writes a one-page playbook before the pilot ships. If there is reliable evidence of exploitation, a named founder files an early warning through ENISA's Single Reporting Platform within 24 hours, a notification within 72 hours and a final report later (Article 14). Ireland's NCSC notes that no pre-registration is needed and warns not to delay the early warning "pending a deep technical analysis" (NCSC). The playbook also covers informing users, which Article 14(8) requires.

Step 10: what goes into the technical file?

The risk assessment, design and vulnerability handling descriptions, SBOM, support period rationale, test reports and the EU declaration of conformity, as listed in Annex VII. As a small enterprise, Brightwell may use the simplified format allowed by Article 33(5). It keeps the file for at least ten years, as Article 13(13) requires.

What does the plan look like by stage?

Stage Security work
Before first PCB Regulatory map, risk assessment v1, MCU with secure boot, key storage, debug lock plan
Prototype firmware Unique credentials, TLS to cloud, signed OTA with A/B, SBOM in CI
Before pilot Disclosure policy, security.txt, reporting playbook, support period decision
Before launch RED technical file, EN 18031 assessment, declaration of conformity, user information
Before December 2027 CRA technical file, Annex I gap check, updated declaration

The Commission's July 2026 guidance includes worked examples aimed at SMEs and is worth reading alongside a plan like this.

Frequently asked questions

When should a hardware startup start on product security?

Before the first PCB is designed. Choices such as secure boot, key storage and debug locking are cheap at design stage and expensive after tooling, and they underpin most regulatory requirements.

Does a seed-stage startup have to comply with the Cyber Resilience Act?

Yes, if it places products with digital elements on the EU market. The CRA has no size exemption, although microenterprises and small enterprises cannot be fined for missing the 24-hour early-warning deadline and may use a simplified technical documentation format.

Are pilot units covered by the regulations?

Units supplied to customers in the course of a commercial activity are generally made available on the market, so the rules apply. Internal prototypes that are never supplied are different. If in doubt about a specific pilot arrangement, take advice.

Is an air-quality sensor an important product under the CRA?

Not on its own, based on the Annex III list, so it would normally be a default product that can self-assess. Adding a function listed in Annex III, such as a security alarm or camera, could change that.

What is the most expensive security decision for a hardware startup?

Usually the support period. Committing to at least five years of security updates means long-term cloud, maintenance and triage costs, which should be built into pricing and funding plans early.

Building a connected product and want a second opinion on its security plan? Book a free 30-minute call with our team.

Cover photo by Matheus Bertelli on Pexels.

Hephaestus Technology

We help founders and businesses validate, build and launch products across software, AI and hardware, from offices in Ho Chi Minh City, Dublin and Delaware. See how we work.

Tell us what
you're building.

A free 30-minute call. You'll get an honest view of the idea, the risks we see and what we'd do first, whether or not we end up working together.