Express + expert help

View pricing →
Skip to main content
tutorialCRA4 min read

CRA Cybersecurity Risk Assessment for an IoT Gateway

Create an IoT gateway risk register that connects product boundaries, abuse cases, security controls and verifiable release evidence.

By Cenitia

A CRA cybersecurity risk assessment should connect the actual product and its intended use to credible abuse scenarios, controls and verification evidence. A generic list of threats with every row marked “low risk” does not explain why the released gateway is appropriately protected.

This walkthrough uses an invented industrial gateway. Its risk register is a worked planning example, not a tested device assessment.

Define the product boundary

The model gateway bridges Ethernet sensors to a manufacturer designed remote management application. It has local commissioning, signed firmware updates and a service port. We assume deployment in a managed industrial network, but assess reasonably foreseeable misuse such as exposed management access and reused credentials.

Draw the interfaces and trust boundaries. Identify who can issue commands, provision credentials, sign updates and access collected data. Separate assumed environmental protections from controls the product itself provides.

CRA Article 13(2) establishes the manufacturer's cybersecurity risk assessment obligation. The July 2026 Commission guidance, section 7, explains assessment and integration of components and external dependencies. The regulation governs; the guidance is non-binding and our register is a suggested implementation aid.

Write abuse cases that can drive design decisions

Model abuse caseExposure to investigateCandidate controlEvidence needed
Attacker reaches commissioning after installationInterface remains remotely reachableLimit commissioning state and require authenticationState transition and network access tests
Attacker substitutes an update packageUpdate input and trust anchorSigned package validation and protected keysPositive and negative signature tests
Compromised backend sends device commandsDevice trusts remote requestsAuthorisation, command bounds and loggingBackend and device permission tests
Service port exposes secretsPhysical or maintenance accessRestricted service design and secret handlingDebug configuration and extraction review
Dependency loses upstream supportPatch availability during expected useMigration or maintained patch planDependency ownership and support decision

The controls are candidates. Writing them into the register does not show they are implemented. Mark the evidence status honestly: proposed, implemented, verified or unresolved.

Explain likelihood and impact

Choose a scoring method that your team can use consistently. Define the meaning of each likelihood and impact level, including the environmental assumptions. Score the scenario before and after the control using the same definitions.

Avoid treating a numerical score as an objective measurement. “Impact 4” means little without explaining the possible harm: unauthorised command execution, loss of availability or exposure of credentials. Also record uncertainty, especially when proprietary components cannot be inspected.

A low estimated likelihood should not erase a severe plausible impact. Ask whether the estimate depends on an installation instruction customers may not follow. Where your control relies on network isolation, document how that requirement is communicated and what happens if it is absent.

Map each scenario and selected control to the relevant CRA Annex I requirements and supporting technical evidence. Explain requirements that are assessed as not applicable or need a different implementation. A standards checklist can support the assessment; it should not replace the product specific reasoning.

Keep component and service dependencies visible. A commercial cloud provider's security statement is useful context but does not prove that your device's command authorisation is correct. Assess the integration, including failures and misuse at the boundary.

Make verification falsifiable

For the commissioning row, define the expected result before testing: an unauthenticated remote request cannot enter commissioning after the product is enrolled. Record the firmware hash, state, network setup, request and observed response.

Test a permitted path too. A negative test that always fails because the interface was broken cannot establish that the authorisation control behaves correctly. Retain raw evidence and distinguish a failed test from an accepted residual risk.

Record the release decision and reassessment triggers

Name the risk owner and approver. Open high consequence gaps need a documented treatment decision, not an empty “accepted” checkbox. Internal risk appetite or commercial cost alone cannot override the CRA essential requirements; the guidance addresses that limit in paragraphs 157–161. Identify triggers such as a new radio option, remote control feature, dependency vulnerability, cloud interface change or support plan revision.

None of the model controls have been tested on a real gateway. The assessment method does not decide the product's CRA classification or conformity route; those require their own review.

CRA applicability and timing

Most CRA product obligations apply from 11 December 2027; manufacturer Article 14 reporting applies from 11 September 2026. Check the product scope, role and specific transitional provisions before treating a future requirement as a current duty. See the Commission manufacturer guidance and reporting guidance.

Continue reading

Related guides

Put this into practice

Free tools & references

New to the terminology? Browse the compliance glossary — plain-English, citation-backed definitions of every term above.