iac.htora.dev · security templates

Home/For responders/Running a cloud incident as an organisation

Running a cloud incident as an organisation

A cloud incident needs four jobs done at once: deciding, doing, writing it down, and telling people. This lesson names who does each, sets the rhythm they work to, and shows how a team grows into it.

1Declarename the commander, open the log2The first hourpreserve, scope, contain, first update3Operational periodsobjectives, a briefing, assignments4Hand overa briefing, announced to everyone5After actioneach finding becomes a change
The roles stay the same from the first minute to the last. What changes is how many people share them.

Where the roles come from

The Incident Command System is the structure the National Incident Management System prescribes. FEMA's doctrine, third edition, October 2017, describes it. Fire services, hospitals, and utilities use it, so the people you call will know it.

The doctrine puts three officers beside the commander: public information, safety, and liaison. It typically puts four sections under the commander: operations, planning, logistics, and finance and administration. Three of its rules carry over to a cloud incident unchanged:

  • A function nobody has taken stays with the supervisor. Until the commander hands out a role, the commander has it.
  • The structure grows only as needed. The doctrine calls this modular organisation: positions are filled as the incident's size and complexity demand.
  • Span of control. The doctrine gives one supervisor to five people as the optimal ratio, and calls it a guideline for judgement.

Five roles for a cloud incident

This site adapts the doctrine to five roles. Logistics and finance usually fall to the commander in a cloud incident: access for responders, a forensics account, and the provider's bill.

RoleFrom the doctrineDoesWorks from
Incident commanderIncident CommanderSets objectives and priorities, approves containment that breaks production, and owns the timeline.The first-hour refresher and the list of containment actions agreed in advance
OperationsOperations SectionThe hands on the cloud consoles and command lines: preserve, scope, contain, eradicate.Evidence that survives termination, the containment refresher, saved queries such as those in insider data theft, and every alert the templates create
Planning, or scribePlanning Section: its Situation and Documentation UnitsKeeps the timeline, the decision log, and the evidence register, and drafts the next period's objectives.The manifest and hashes the evidence template writes, checked with evidence-verify
CommunicationsPublic Information OfficerWrites updates for leadership, staff, and customers, and drafts any notice with Legal.The four-part update in Leading a cloud incident
Liaison, where neededLiaison OfficerThe single contact for outside parties: the cloud provider's support, an outside response firm, insurers, and law enforcement.A contact list written before the incident, with account and contract numbers

Operations needs the cloud lessons at hand: AWS, Azure, Google Cloud, Oracle Cloud, and Kubernetes.

From one responder to a full team

PeopleWho has which roleWhat to watch
OneEvery role. The responder commands, works the consoles, keeps the log, and sends updates.The log and the update schedule are the first things dropped. Keep both, even if each entry is one line.
TwoA commander, who also scribes and writes updates. An operations lead.The commander stays off the consoles.
Three or fourAdd a scribe, then communications.Updates go out from one person, so leadership hears one account.
Five or moreAdd liaison. Split operations by cloud or by task, each part with a lead.Keep each lead near five people, the doctrine's guideline.

Agree in advance what makes a lone responder call for help. Good triggers are ones anyone can check: a second account or cloud is involved, containment would break production, or personal data may be exposed.

Handing over command

The doctrine says a transfer of command includes a briefing that captures what the next commander needs, and that everyone involved is told. It also names ICS Form 201, the incident briefing, as the form a first commander typically fills in before formal planning starts.

For a cloud incident, the outgoing commander covers these, out loud and in the log:

  • The current objectives, and which are met.
  • What is contained, and what is still open.
  • Decisions waiting, and who is waiting on them.
  • Who has which role, and the next update and briefing times.
  • Where the decision log and the evidence register are.

The new commander then posts one line in the incident channel naming themselves and the time. Until that line exists, the outgoing commander is still in command.

The first hour

  1. Name the commander, and say so in the incident channel.
  2. Open the decision log and the timeline. Record when the alert or report arrived.
  3. Preserve: tag affected instances for the evidence template, and export logs close to their retention limit.
  4. Scope: search the audit logs for new keys, identities, and trust changes.
  5. Contain with the actions agreed in advance.
  6. Send the first update, with the time of the next one.
  7. Ask whether personal data, a regulated service, or a public company's material systems could be involved. If any could, call Legal now.

The first-hour refresher puts the steps on one page.

The operational period

The doctrine defines an operational period as the time scheduled for a set of actions in the incident action plan. It says those plans generally cover 12 to 24 hours. Pick a length that matches your shifts.

Each period starts with a briefing. In the doctrine, command and the sections present the objectives, review the situation, and share communications and safety information. A cloud briefing covers the same ground:

  • Objectives anyone can check. "Every key the compromised user created since the first event is disabled" can be checked. "Contain the attacker" cannot.
  • The situation: what changed since the last briefing.
  • Assignments by name, one owner each.
  • Decisions taken since the last briefing, read from the log.
  • The update schedule and the next briefing time.

Write the plan on one page. The doctrine calls this management by objectives: specific, measurable objectives, with results recorded against them.

Decision records

The doctrine's Documentation Unit keeps incident files for legal, analytical, and historical purposes. The decision log is that file. Each entry has seven fields:

  • The time, in UTC.
  • The decision, in one sentence.
  • Who made it.
  • The evidence it rested on.
  • The options set aside.
  • Who carries it out.
  • When to look at it again.

NIST SP 800-61r3 lists recording investigation actions, with the records' integrity and provenance kept, as RS.AN-06. The GDPR, in paragraph five of Article 33, requires a controller to document every personal data breach: its facts, its effects, and the remedial action taken.

This page gives no legal advice. It quotes two primary texts so the commander knows why Legal is called early. Which rules apply is for counsel to decide.

TextWhat it says about timeWhat the clock runs from
SEC Form 8-K Item 1.05, for registrantsA filing within four business days of the registrant determining that a cybersecurity incident is material. The determination is made without unreasonable delay.The materiality determination. The SEC allows a delay when the US Attorney General finds disclosure poses a substantial risk to national security or public safety.
GDPR Article 33, for controllersNotice to the supervisory authority without undue delay and, where feasible, within 72 hours. A processor tells the controller without undue delay.Becoming aware of the breach. No notice is needed if the breach is unlikely to risk people's rights and freedoms. Details may follow in phases.

GDPR Article 34 adds telling the people affected, without undue delay, when the risk to them is likely to be high. Contracts, sector rules, and other laws add their own clocks. Counsel keeps that list.

What the commander can do is record the two moments these texts start from: when the organisation became aware, and when any materiality decision was made. Both go in the decision log, with who was present.

Call Legal when any of these may be true: personal data is involved, an SEC filer faces a material incident, an insider is involved, or law enforcement is to be contacted. Insider cases have their own lesson.

Phases, mapped to NIST SP 800-61r3

NIST published SP 800-61r3 in April 2025. It replaces SP 800-61r2, from August 2012, and is written as a Community Profile of the Cybersecurity Framework 2.0.

The second revision had a four-phase life cycle: preparation; detection and analysis; containment, eradication, and recovery; and post-incident activity. The third revision builds its model on the framework's six functions instead.

  • Govern, Identify, and Protect are preparation. NIST places them outside the response itself.
  • Detect, Respond, and Recover are the response.
  • Lessons from every function feed the Improvement category of Identify, ID.IM.

The third revision also drops detailed procedures. NIST says they change too often for one static document, and keeps its links on a website it can update. It recommends documenting every incident response role in policy, as GV.RR-02.R1.

This site's phaseFramework 2.0Who leads it
DetectDetect; then RS.MA-02, reports triaged and validatedWhoever is on call, until a commander is named
Command and updatesRS.MA, incident management; RS.CO, reporting and communicationCommander and communications
Scope and preserveRS.AN, incident analysis, including RS.AN-07 on keeping evidence's integrity and provenanceOperations, with planning recording
Contain and eradicateRS.MI-01, contained; RS.MI-02, eradicatedOperations
RecoverRC.RP, recovery plan execution; RC.CO, recovery communicationOperations, with the system owners
ReviewID.IM, improvementCommander, with the owning teams

A maturity ladder

This ladder is the site's own. No published model stands behind it. Each rung is met when you can show the thing it names, and a rung counts only if the ones below it are met.

RungMet when you can show
1. ReachableAn on-call rota with a named backup for every shift, and a test page from the last quarter that reached the person on call.
2. Roles writtenA role card for each of the five roles, a list of who may command, and the containment actions agreed in advance, signed off by the system owners.
3. RehearsedExercises at a frequency the plan states, with dates on record. The evidence template has been run against a test instance on the same schedule.
4. TrackedEvery after-action item has an owner and a due date in the tracker the owning team uses, and the open count is reported.
5. ImprovingEach closed item names the change it made: a template, a detection, a role card, or a refresher. The next exercise tests that change.

A small team can reach the third rung with one person on call. The ladder measures what is written and rehearsed, and headcount does not move it.

After action

Run the review within a week, while people remember. Record the timeline from the first attacker action, the time each phase started, the decisions from the log, and what slowed each phase down.

A review is finished when each finding has become a change, owned by the team that owns the system:

FindingThe change
Nobody was clearly in command at firstA commander rota, and naming the commander as step one on the first-hour refresher
Containment waited for approvalAdd the action to the list agreed in advance, signed off by the system owner
An alert fired and nobody knew what it meantLink its entry on the alerts page from the runbook
Evidence was lost when the instance was replacedEvidence that survives termination
The attacker deleted or disabled the logsAudit logs nobody can delete
Legal heard lateThe Legal triggers on the communications role card
A decision was lost at handoverThe handover list read out, and the new commander's line in the channel

The incident command refresher puts the roles and the rhythm on one printable page.

Sources

  • FEMA, National Incident Management System, third edition, October 2017. The roles, span of control, transfer of command, operational periods, and the planning cycle.
  • NIST SP 800-61r3, April 2025. The mapping to the Cybersecurity Framework 2.0 and its Table 1.
  • SEC, Release No. 33-11216, 26 July 2023, and Form 8-K Item 1.05. The four business days, the materiality determination, and the Attorney General delay.
  • Regulation (EU) 2016/679, the GDPR, 27 April 2016. Articles 33 and 34.

Registry 0.23.0. Built 2026-11-17.

Made by Habibullah Tora. Source on GitHub. Code under the MIT licence, writing under CC BY 4.0.