Administrative teams behind patient-facing websites and booking systems sometimes ask how they can tell whether their security habits are working or merely busy. This checklist turns ThreatOps into twenty-four questions you can answer yes or no in about an hour. It concerns office systems, logins and vendor feeds only; it says nothing about clinical matters and is not advice of a medical or legal nature. Print it, assign an owner and score honestly.
How to use the audit
Work through each section with the person who actually does the task, not their manager. Mark “yes” only if you can show evidence, such as a document, a ticket or a log entry. Every “no” becomes a line in a gap list, which is a respectable output in itself. The method is based on practices described in the Ops Atlas page about ThreatOps, which defines the discipline as intelligence gathering and hunting run together and admits that no formal standard covers the term.
Section A: Questions and ownership
- A named person owns threat intelligence and hunting, with time set aside weekly.
- A written list of priority questions exists, with roughly five to eight open at once.
- Each question carries one fact needed and one decision it will inform.
- The list was reviewed within the last quarter, and unused questions were removed.
If you scored fewer than three yes answers here, stop and fix this section first. Without questions, everything below is decoration. When your technical staff are still building skills, a practice plan for working through DevOps interview rounds is one low-cost way to structure their learning.
Section B: Sources and trust
- A register lists every source you rely on, including free ones.
- Each source has a reliability letter and each claim a credibility number, using the FIRST scheme the Atlas describes.
- You can name which feed answers which question; unmatched feeds are marked for review.
- Someone has calculated each feed’s precision, meaning the share of its alerts that proved real.
For reference, the Atlas’s worked example shows a feed with 18 genuine alerts out of 120, a precision of 15 percent, which kept it on alert-only duty and flagged it for cancellation. If you have never computed this number, that is your first gap.
Section C: Expiry and blocking
- Every indicator carries an end date, set by you, because the exchange format leaves expiry optional.
- Different indicator types have different lifetimes, written down.
- No automatic block exists unless the source grade, claim credibility, measured precision and allowlist checks were all satisfied.
- Anyone on call can reverse a block, and the owner reviews reversals promptly.
- A monthly count of reversed blocks is kept.
The Atlas’s sample gates require precision of at least 0.9 over ninety days. Whatever threshold you choose, the discipline is to measure before you automate. Appointment-style preparation offers a good parallel: tips for getting ready for a checkup visit make the point that a little preparation beforehand makes the visit itself far more productive, and the same holds for preparing your blocking rules.
Section D: Hunting
- A hunt calendar exists, with a rhythm close to one hunt every two weeks.
- Each hunt states a hypothesis covering actor, behaviour, location and evidence.
- Each hunt tests one or two behaviours only.
- Log history covers at least ninety days for the systems under test.
- Every completed hunt produced a written output, for instance a rule request, a log-coverage fix, an escalation to the response team or a confirmed-negative note.
- A colleague who did not take part could rerun the hunt from the notes.
Be strict about the last two. The Atlas treats a null result as a success when the data was adequate, and a gap ticket as the correct output when it was not. Hunts that end in conversation only should be scored “no”.
Section E: Sharing and reporting
- A written sharing policy names permitted categories and applies colour labels to anything leaving the team.
- Staff know not to forward amber-labelled items beyond the intended recipients.
- Regulatory sharing options have been read; the Atlas mentions the voluntary article in the EU’s NIS2 directive and notes that the US sharing protections run to 11 December 2026, so current status should be checked.
- Reports state probability with standard words and confidence separately.
- Quarterly reporting uses ratios, not raw counts of indicators, reports or hunts.
Evidence to gather before you score
Auditors trust paper, so collect it in advance. Pull the priority-question list, the source register, a sample of indicator records showing their end dates, the last three hunt write-ups and the sharing policy. Ask for the log retention settings for the booking portal, the staff login system and the email gateway. If any item takes more than a few minutes to find, that difficulty is itself a finding, because nobody can maintain what nobody can locate.
It also helps to hold the audit meeting away from the usual desks, with a facilitator who does not own the answers. People are more candid about gaps when the exercise is framed as a health check on the process rather than on individuals.
Reading your score
Add up your yes answers out of twenty-four. Fewer than eight means you are essentially consuming feeds, the stage the Atlas places near the bottom of its ladder, where feeds should stay in alert-only mode and expire. Eight to sixteen suggests a question-driven programme, with regular hunting as the next goal. Above sixteen, you are nearing the upper steps of the Atlas ladder, where hunts run on a schedule and intelligence guides priorities.
Turn each “no” into a dated task, and pick no more than three for the coming month. The Atlas’s thirty-sixty-ninety outline is a handy order of attack: ownership, five questions, a source register, expiry and a sharing community first; templates and two hunts next, aiming for a first accepted detection request; then tooling and a decision about renewing or cancelling one feed on evidence.
Finally, keep the results. Repeat the audit each quarter and compare scores. A modest, steady improvement tells you more than a single high mark ever could.