Security
Cybersecurity tabletop exercise: how to design one that finds something
A tabletop only finds something if the room can fail it. Here is how to design one that can, and who has to be in it.
Computing America Cyber. Security assessments, penetration testing and staff training.
Send the plan
In short
- A tabletop is a discussion. It won't tell you whether backups restore, detection fires or the phone tree works, and one that claims to have validated those is overclaiming.
- CISA publishes over a hundred downloadable tabletop exercise packages, including ransomware and industrial control system compromise. Scenario design is not the scarce input.
- Two questions produce most of the value: who decides to stop production, and who decides whether to pay. Both are business decisions that turn out, mid-incident, to have no owner.
- Ask "which log would show that?" at every step. A missing source is a finding about the estate, not a fault of the analyst.
- The deliverable is a short list of decisions and evidence gaps, each with a named owner and a date.
Take a CISA package and add injects that remove what your plan assumes: the decider is unreachable, the backup is encrypted, the log is past retention. Otherwise a room of competent people will find the answer the plan already handles. The rest of the work is the two decisions that stall a room, the people who get called, and a deliverable with owners and dates.
An exercise that finds nothing looks like this: a scenario is read out, people describe what they would do, the room agrees it was valuable, and a report circulates the next week. Nothing changes, and it repeats annually because a policy or an insurer requires it. An exercise the room can't fail has measured nothing.
The scenario is already written¶
CISA publishes over a hundred tabletop exercise packages (opens in a new tab), covering ransomware, insider threat, phishing and industrial control system compromise, with versions for water systems, healthcare, local government, ports and elections. Each has objectives, a scenario, discussion questions, slides, feedback forms and an after-action template. What's scarce is a facilitator who pushes where the room gets uncomfortable, and an organization willing to reach the decision it has been avoiding.
Ask the two questions as hard versions¶
Neither question is technical, and both stall a room that believed it had a plan.
- Who decides to stop production? In a plant or utility, the containment step that limits damage can halt something earning money or delivering a service, and that isn't the security function's call. Ask who has the authority, then ask what happens if that person is on a plane.
- Who decides whether to pay? It's easy to keep hypothetical, and that is how it becomes a 3am decision by whoever is awake. "What would we do?" invites a principled answer that binds no one. Ask who signs, what the threshold is, who must be told first, and whether the insurer or counsel must approve.
Put the people who get called in the room¶
An exercise attended only by security tests the group that already knows the plan. The people who decide whether a response works are operations, the plant manager, legal, finance, communications and whoever owns the customer relationships. In an OT environment, add the integrator: ask whether they answer the phone on a weekend, and what the contract says.
Test what only your own estate can answer¶
A tabletop works on two layers a generic range can't reach, evidence and decisions. Skills such as reading a capture, pivoting and querying logs are a third layer, and a generic training range is the right purchase for those.
| Layer | Example question | Who answers |
|---|---|---|
| Skills | Can the analyst read this capture? | A trained person, on any range. |
| Evidence | Do the logs that would show lateral movement exist, and for how long are they kept? | Your systems and the people who run them. |
| Decisions | Who can pull a segment off the network at 2am, and who stops production? | A named owner in your organization. |
One inject that tests the evidence layer: at T+40 minutes the backup repository is encrypted. Who confirms it? Which log shows when it started? The expected evidence is the date of the last good restore.
Logging can fail quietly: sources not collected, a verbosity that omits the field that mattered, thirty days kept when the intruder sat in for ninety. Ask "which log would show that?" at every step, and record a missing source as a finding about the estate. To test whether your logging records a specific action, perform the action on a test host, such as creating an account or reading a share, and look for it in the log store. That is a technical test, not a tabletop, and it's worth running once the tabletop has told you which logs to doubt.
days before the exercise
0 Exercise day
Log store holds the last 30 days
0–30 Logs kept: 30 days
30 Oldest log still kept
30–90 No log to read: 60 days
90 The intruder's first entry
The log that would show it is gone
Thirty days of log retention against a ninety-day intrusion leave the first sixty days with no log to read.
SourceExample scenario from this article, not measured data.
Leave with owners and dates¶
People describe what they would do better than they do it, so treat answers as claims to check. Training the staff who will be the first to see the problem is a separate job, which role-based exercises and phishing simulations cover.
The deliverable is a short list with a name and a date on each. Two may be the questions above. One will be a piece of evidence the room couldn't produce. That is where an exercise becomes an engineering change: if no one can say which accounts touched a share, the finding is a logging gap and it goes in a backlog. Treat each item like any other defect, and book the date you will check it closed.
Sources
Next step
Send us your incident plan.
We will send back the six injects it is least prepared for, drawn from the plan's own gaps, so you can run the exercise yourself whether or not you run it with us.
- Phone
- (214) 723-2510
- Reply
- A person replies, not a sequence: within one business day, from someone who would be on the engagement.
Packages
Start with a package
Four packages, each with its price on the page. Pick one and buy it online.
$199/month
Billed monthly
$182.42/month
Billed yearly
$499/month
Billed monthly
$457.42/month
Billed yearly
- Scammers stopped from sending email as you
- Your website and systems scanned every week
- Your email settings checked every week
- Everything in the 1 to 10 plan
- Covers up to 50 people
$199/month
after a one-time setup fee
- A site planned, designed and built for you
- Hosting and updates every month
$499/month
Billed monthly
$457.42/month
Billed yearly
$999/month
Billed monthly
$915.75/month
Billed yearly
- Your Google profile, fully set up
- A Google post every week
- 2 social posts a week
- Everything in Get found
- 4 social posts a week
- Your top 10 Google searches tracked
from$500/month
after a one-time setup fee
- Your data gathered from the tools you use
- A first analysis that answers your questions
- New findings every month
Related reading
Engineering4 min readForensic readiness: the logging you need before a breach
Retention, endpoint telemetry and audit trails are decided months before an incident, and they set what an investigation can ever establish.
Security5 min readOT network segmentation: when the drawing and the switch disagree
The plant network diagram shows a segmented plant. The switch can show a permissive VLAN rule, a two-homed workstation, and a vendor's remote path that isn't on the drawing.
Security4 min readVulnerability research: most of the work is ruling things out
A recon pass hands you hundreds of finding-shaped candidates. Which ones die is a person's call; two gates around that call are mechanical.
