TL;DR.

A status change is not closure. An automotive incident is closed only when the confirmed scope, a bounded technical explanation, containment and remediation, matching verification, communication decisions and the residual risk are recorded and approved by a role with the authority to accept that risk.

"Closed" is one of the most ambiguous words in an incident system. It may mean that an analyst completed a task. It may mean that a supplier delivered a patch. It may mean that the alert which opened the case stopped firing. None of those conditions proves that the automotive risk has been fully assessed, treated, verified and approved. This automotive incident closure checklist is written for incident managers, PSIRT leads, CSMS auditors and compliance owners who need "closed" to mean something a reviewer can test.

An evidence-backed closure package should let a reviewer, months later and without the original participants in the room, understand what happened, what was affected, what the organization decided, and why the remaining risk is acceptable. The sections below walk the closure decision from scope through approval, assemble the package as a checklist, and apply it to a worked example in which a validated patch was mistaken for a closed incident.

Why a status change is not an incident closure

Most incident tools model closure as a state transition: someone moves the record to "closed" and the dashboard count goes down. That is fine for a help desk. It is not fine for a cybersecurity management system, where the closure record is the argument an auditor, a type-approval authority or a product-liability lawyer will read years later.

An automotive incident record has to show that the organization understood the affected population, chose a treatment proportionate to the risk, verified that the treatment did what it claimed, decided who had to be told, and accepted the remaining risk through a role entitled to accept it. ISO/SAE 21434 and UNECE R155 expect incidents to be handled as part of the product's continual cybersecurity activities; neither prescribes a closure form, so each program must define its own criteria.

A closed incident is a decision with evidence behind it, not a status with a timestamp.

Closure begins with a stable scope

The closure record should identify the confirmed affected scope and explain any changes made to it during the investigation. Scope may include vehicle programs, variants, ECUs, hardware versions, software builds, third-party components, suppliers, customer configurations and markets.

The record should also capture the basis for exclusions. If a vehicle variant is declared not affected, the decision needs supporting evidence, such as:

  • A build configuration showing the vulnerable component or function was never compiled in for that variant.
  • A component version, confirmed from the SBOM or the supplier's declaration, outside the vulnerable range.
  • A disabled function, verified by variant coding or production configuration rather than by assumption.
  • An unreachable interface, demonstrated by the network communication matrix or a test from the relevant attack surface.
  • A verified compensating control already in place, with a reference to the test that verified it.

Consider a worked case. A PSIRT receives a report affecting a Bluetooth stack used in, say, 14 variants across three programs. Nine are excluded because the affected profile is not enabled in their variant coding, two because they ship a later stack version, and three remain in scope. Each exclusion is attached to the record that supports it, so that when a new report arrives against the same stack, the team re-runs the question against current SBOM records instead of rebuilding the analysis from memory.

This distinction prevents a common failure: closing the known case while leaving a related program unexamined. A record that lists only the vehicles found affected cannot tell a reviewer whether the rest of the population was excluded or forgotten.

Preserve a bounded technical explanation

Not every investigation produces a perfect root cause. The closure package should still provide a bounded technical explanation supported by the available evidence, and it should be honest about where the boundary sits.

The record may include the initial signal, reproduction steps, relevant logs, the affected code or component, exploit prerequisites, the attack path, the TARA scenarios the incident maps to, and the conditions under which impact can occur. Where uncertainty remains, state it and show how the treatment addresses it.

A vague conclusion such as "unable to reproduce" is not a sufficient closure rationale. The team should say what was tested, under which configuration and build identifier, with which result, and why that result changes the decision. "Unable to reproduce on bench build 4.2.1 with the production calibration across the three reported trigger sequences; the reporter ran a pre-series build not in the field" is a rationale. "Could not reproduce, closing" is not.

The technical explanation is also where the incident meets the risk assessment. If the incident realizes a scenario already in the TARA, the record should reference it and note whether the assumed feasibility and impact held. If it reveals a path the TARA did not consider, that is itself a finding, and the TARA revision it triggers belongs in the closure package.

Separate containment from remediation

Containment reduces immediate exposure. Remediation addresses the underlying issue. Automotive incidents often need both, frequently months apart.

A temporary intrusion-detection rule, a blocked diagnostic service, a disabled feature, a certificate revocation or an operational restriction may contain the threat while engineering prepares a software update. The closure record should make clear whether each action is temporary, which population it covers, what condition allows its removal, and what follow-up remains open.

If the organization closes the incident before permanent remediation, it should retain a linked action with an owner, a deadline, a deployment scope and an escalation rule for when the deadline passes. Closure should not make unfinished treatment invisible. Linking the open action to the incident, and the incident to the affected population, keeps that population visible in the vulnerability management lifecycle after the incident has left the daily agenda.

Verification must match the claim

Evidence should prove the specific action claimed by the closure decision. This is where closure packages most often fail quietly.

If the team claims that a patch removes a vulnerable code path, testing should exercise that path; a regression suite that never calls the affected function proves the patch broke nothing else, not that the vulnerability is gone. If the team claims that network segmentation prevents reachability, the validation should test the relevant route on the production gateway configuration, not a development bench with permissive rules. If the team relies on a supplier statement, the statement should name the affected product and version and provide a verifiable justification; the OEM–supplier vulnerability disclosure protocol describes what a usable statement contains.

Closure claimEvidence that matchesEvidence that does not
Patch removes the vulnerable code pathTest exercising the path on the corrected build, with build identifier and resultGeneric regression pass; supplier release note alone
Segmentation prevents reachabilityRoute and firewall test from the attacker's position on production configurationArchitecture diagram; bench test with development rules
Feature disabled on affected variantsVariant-coding read-back from production configuration per variantEngineering intent note; configuration spreadsheet
Supplier component not affectedSupplier statement naming product, version and technical basis, checked against the SBOMEmail saying "not affected" without version or reason
Compensating control in placeReference to the control, its current test result and the population it coversControl listed in the TARA without a current test

For each verification the package should record the test environment, build identifier, procedure, result, reviewer and the link to the control or mitigation being verified.

Record communication and reporting decisions

Automotive incidents may trigger customer, supplier, regulator, insurer or internal governance communication. The closure package should show which notification questions were assessed, who made each determination, and what submissions or messages were sent. A decision not to notify needs the same record.

The record should also preserve the time basis. For regulations with reporting windows, such as the CRA obligations that began on 11 September 2026 for covered products, the team must show when awareness occurred and how deadline decisions were handled. The first external report, the first internal triage and the moment the PSIRT confirmed the product was affected can be days apart; the record should state which event started the clock and why, so the timeline can be defended rather than reconstructed under pressure.

In a multi-tier incident, the package should show the escalation path used, the requests made, and the responses received with their dates. The OEM–supplier incident escalation framework sets out the roles behind those entries; for fleet-facing messages, the fleet-wide incident response playbooks cover what customers and operators need to hear.

Require authorized approval of residual risk

Closure changes the organization's risk position. The person approving it should have defined authority for that decision, proportionate to the residual risk being accepted. An analyst should not be able to close a case that leaves, say, 3,200 vehicles unpatched under a compensating control, because that is a risk-acceptance decision for a named product-security or program role.

The approval record should contain the selected treatment, the residual risk statement, the open actions and their owners, the evidence the approver reviewed, the decision date and the rationale. Where multiple functions hold decision rights, the workflow should capture each required approval rather than collapsing them into one generic owner: a technical approval from the PSIRT lead on the engineering argument, a program approval on deployment and customer position, and, where a submission was made, a compliance approval on the reporting assessment.

Residual risk is accepted by a role, not by whoever happened to have the ticket open.

Approval authority should be defined before the incident, in the CSMS process, not negotiated during it.

The automotive incident closure checklist

A complete package should contain the items below. The third column names the record each item normally comes from.

Checklist itemWhat it must showTypical source record
Incident source, awareness time, classification and chronologyWho reported, when the clock started, how severity was set, the sequence of eventsIncident record, intake log
Confirmed affected and excluded product scopePrograms, variants, ECUs, builds and components in scope; evidence for each exclusionProduct structure, SBOM, variant configuration
Technical analysis and relevant risk contextBounded explanation, attack path, prerequisites, linked TARA scenariosInvestigation notes, TARA
Containment and remediation actionsEach action, its population, temporary or permanent, removal conditionChange and deployment records
Validation evidence tied to each treatment claimEnvironment, build, procedure, result, reviewer, linked controlTest records
Supplier and customer communicationsRequests, responses, notifications, datesSupplier case records, communication log
Regulatory reporting assessment and submissions where applicableQuestions assessed, determination, submissions, time basisCompliance assessment
Updated TARA, SBOM, vulnerability, control or cybersecurity case recordsWhich records changed, revision referencesLinked engineering records
Residual risk statement and authorized approvalRisk accepted, by whom, on what evidence, on what dateApproval workflow
Lessons, control improvements and linked follow-up actionsWhat changes, owner, deadline, escalation ruleAction register

This structure gives incident closure a clear test. If a reviewer cannot trace the decision from signal to evidence to authority, the incident is not ready to close. It is the same test an auditor applies when sampling closed incidents during a CSMS audit.

Worked example: a validated patch does not prove fleet closure

An engineering team reproduces a vulnerability in a gateway diagnostic handler and validates a corrected build on the reference test bench. That is strong remediation evidence. It is not proof that the incident can close.

Walk the checklist. Scope: the corrected build applies to one hardware revision, and the team has not confirmed whether the second revision, built on a different supplier line, shares the handler. Treatment: the fix exists, but the deployment plan is unapproved and, say, 2,000 units in a market without over-the-air coverage will need a workshop visit. Communication: the customer-notification question for that market has not been assessed, and the CRA reporting determination reads "pending" with no owner. Approval: the only signature on the record is the test engineer's.

The closure gate should distinguish technical correction from operational completion. Three explicit states help:

  1. Active investigation: signal received, scope not yet stable.
  2. Controlled remediation: scope confirmed, containment in place, fix validated on the corrected build; deployment, communications and residual-risk approval still open.
  3. Closed: every checklist item complete and approved by the authorized role; only linked follow-up actions with owners and dates remain.

Here the case moves to controlled remediation on the strength of the validated build and the containment rule on the diagnostic service. Final closure occurs only when the defined scope, treatment, verification, communication and residual-risk conditions are satisfied and approved by the authorized role: after the second hardware revision is confirmed, the workshop campaign is approved with a deadline, the notification decision is recorded, and the product-security lead accepts the residual risk for units not yet reached.

How a connected incident record supports closure

Every item in the checklist is a relationship: an incident affects a product object; the object carries a build; the build contains a component; the component maps to a vulnerability; the vulnerability realizes a TARA scenario; a control mitigates it; a test verifies the control; a supplier owns the component; an approval applies to a residual risk. A package assembled from documents restates each relationship by hand, and every restatement can drift.

A connected record makes the closure argument a query rather than a reconstruction. A reviewer can start at the approval and ask which evidence it rested on, which tests produced that evidence, which build those tests ran against, and which variants that build reached. ThreatZ supports this by keeping the incident linked to the affected product objects, TARA scenarios, vulnerabilities, controls, tests, evidence, suppliers and approvals, as described on the ThreatZ operations page.

The practical test is cheap to run. Take one recently closed incident and attempt to answer the ten checklist items from the record alone. If the team must call the original participants to reconstruct the answer, the closure record is incomplete, whatever the status field says. Comparing a built-for-purpose record against your current setup helps decide what to change once that gap is visible.

Frequently asked questions

What is the difference between containment and closure?

Containment reduces immediate exposure; closure confirms that the incident has met its defined completion criteria. An incident can be contained while permanent remediation, fleet deployment, regulatory reporting or follow-up actions remain open. The record should state which containment actions are temporary and when they will be removed.

Can an incident close when some vehicles are not patched?

Potentially, but only under an authorized and documented decision. The record should identify the remaining population, compensating controls, monitoring, deployment constraints, residual risk, owner and deadline. Changing a status without that evidence is not defensible closure. The approving role must have authority to accept the residual risk for the unpatched population, and the linked action must stay visible after closure.

What evidence should be attached to the closure decision?

Evidence may include confirmed configuration scope, root-cause analysis, patch or containment records, test results with build identifiers, supplier statements naming product and version, deployment status, TARA updates, the communications assessment, residual-risk acceptance and formal approval. Each item should link to the claim it supports; evidence attached without a claim forces every reader to re-interpret it.

Who should approve automotive incident closure?

A role with defined authority for the residual risk being accepted, named in the CSMS process before the incident occurs. Where residual risk remains, program and compliance roles typically hold separate decision rights, and the workflow should capture each approval on its own question rather than one generic owner signing "closed".

Does closure have to wait for a confirmed root cause?

No, but it has to wait for a bounded technical explanation. The record should state what was tested, on which configuration and build, with which result, and why the chosen treatment addresses the impact even where the exact trigger was not isolated. "Unable to reproduce" alone is not a closure rationale. Where uncertainty remains, the residual-risk statement should name it so the approver accepts a known uncertainty.

Key takeaways
  • Status is not closure: a closed incident is an argument with evidence and authority behind it, not a state transition with a timestamp.
  • Stabilize scope first: record the confirmed population and the evidence for every exclusion, so related programs are ruled out rather than forgotten.
  • Bound the explanation: state what was tested, on which build, with which result; "unable to reproduce" is not a rationale.
  • Keep containment visible: temporary actions need a population, a removal condition and a linked follow-up with owner and deadline.
  • Match evidence to claim: every verification must exercise the specific claim it supports, on production-representative configuration.
  • Authorize residual risk: closure is approved by a role with defined authority, with each required function signing its own question.
  • Test one closed case: if answering the ten checklist items requires calling the original participants, the record is incomplete.

Authoritative references

  • ISO/SAE 21434:2021 Road vehicles — Cybersecurity engineering — ISO/SAE, 2021
  • UN Regulation No. 155 Cyber security and cyber security management system — UNECE, R155
  • Regulation (EU) 2024/2847 Cyber Resilience Act, reporting obligations — European Commission, 2024
  • ISO/PAS 5112:2022 Guidelines for auditing cybersecurity engineering — ISO, 2022
  • ISO/IEC 27035-1 Information security incident management, principles and process — ISO/IEC, 27035-1