What follows is an invented story, assembled from general practice, about a mid-sized flight academy and the small team that looks after its systems. No real school is described. The story is useful because it shows how a ThreatOps routine behaves under pressure, particularly the moment a well-meant automation backfires. The practices it illustrates are laid out in the Atlas entry for ThreatOps maintained by xOps School, and I have kept to that source for every factual point.
The setup
The academy trains a few hundred students a year. Booking, scheduling, student records and the maintenance tracker all live in online systems, and three people share the technology work: an operations lead, a systems administrator and a part-time analyst who also teaches ground school. For their first two years, the group treated security as a matter of subscribing to things. They had signed up for two warning feeds, and one of them was connected directly to the firewall so that any address on the list was blocked automatically.
It felt modern. Each morning a report showed how many items had been ingested and how many had been blocked, and the numbers were big. Nobody asked whether the numbers meant anything.
The Tuesday that went wrong
On a busy Tuesday, three students phoned the front desk because the booking portal would not load. A cross-country flight had to be rescheduled, and an examiner was left waiting. The administrator traced the problem to an automatic block. One of the feeds had added an address belonging to a large shared hosting network, the kind of infrastructure thousands of ordinary sites sit behind. The booking portal’s own provider used part of that network, so legitimate traffic was being refused.
The Atlas describes this exact situation as an anti-pattern: a feed blocking a shared content network. Its recommended runbook is short and clear. Reverse the block in minutes. Put the offending feed into a mode where it merely raises flags within the hour. Then keep a monthly count of blocks that had to be undone. The administrator did the first two things within the hour, partly by luck and partly because the firewall made reversals easy.
The review that followed
The analyst asked a question nobody had asked before: how often had this feed been right? Digging through three months of logs, she found that of 120 items that had produced a hit, only 18 were genuinely hostile. That is 15 percent. The team had been so impressed by the big counts that nobody had calculated precision.
The Atlas contains a nearly identical worked example, and the conclusion there matched the one the academy reached: the feed should stay in alert-only mode, and it should be considered for cancellation. Automatic blocking, the document explains, belongs only to sources that have earned it. Its sample rules ask for the top reliability grades, a measured precision of at least 90 percent over 90 days, and no overlap with a list of known-good entries. They also let anyone on duty reverse a block, with a lead reviewing it within an hour.
The operations lead drew up the same sort of policy on one page and taped it above the server cabinet. Parts of it were simple. Any engineer could undo a block at any hour. Any block would be reviewed the next morning. And a monthly line in the operations log would show how many blocks had been reversed.
From plumbing to purpose
The incident did something useful: it made the team ask what the routine was for. They sat down and wrote priority questions. The Atlas advises five to eight, each tied to a single decision, and the academy produced six. Examples included whether student accounts could be taken over, whether the maintenance tracker could be altered without trace, and what the team would do if the booking provider reported a problem on its own side.
For each question they named who would care about the answer. The academy’s chief instructor wanted to know about student records. The accountable manager cared about the maintenance tracker. Having a named stakeholder per question turned the list into something with an audience.
The story would be tidy if the team had stopped there, but the next move was the most valuable one. They began a fortnightly hunt. The analyst, who understood how instructors actually used the booking system, wrote a hypothesis about sign-ins from unusual places. She limited the hunt to a single behaviour, as the Atlas recommends, and defined in advance which records would count as evidence.
The first hunt
The first hunt did not find an intruder. It found something that mattered almost as much: sign-in records were kept for only three weeks, far short of the roughly 90 days the Atlas treats as a minimum for useful hunting. The analyst wrote it up as a gap, not a failure, and the administrator extended retention that month.
Two weeks later the second hunt ran against the longer records. Again nothing hostile turned up, but this time the data was complete, so the analyst could write the result in the stronger form: the records were good enough to show such activity, and none appeared. The Atlas calls this a confident negative and counts it as a success. The write-up was short enough that the part-time ground-school colleague could repeat the check without help, which is the acceptance test the Atlas proposes.
The debrief
Aviation people debrief everything, so the team ran a debrief on the Tuesday outage. They used a plain structure: what happened, why it happened, what we will change, and what we will watch. The “why” section was the interesting one. The outage was not caused by a malicious actor. It was caused by trusting a source more than its record justified, and by having no routine for checking that trust.
The analyst added one more item that the Atlas would approve of. She asked who had been allowed to turn on automatic blocking in the first place, and discovered nobody had formally decided it. A tick-box in the firewall console had been enough. From then on, switching a source from flagging to enforcing required a short written justification with its measured precision attached, reviewed by a second person. It added five minutes to a task and removed an entire class of surprise.
There was also a lighter moment. A student who had been caught by the outage joked that the academy had finally found a hazard that could cancel a lesson on a clear day. The team took it as a reminder that digital systems are now part of the training environment, much like a runway or a briefing room.
What changed in the numbers
After a quarter, the operations lead stopped reporting how much had been ingested and started reporting ratios. Five of six priority questions had a named stakeholder. Of the intelligence items that reached the team, a minority led to something being done, and the lead was comfortable with that because the number was now visible. Of the hunts run, some produced durable changes and some produced notes. Gaps were counted too, with a column for how many had been closed.
Nothing here needed a new product. The academy reached what the Atlas calls the requirements-driven stage largely by being clearer about its questions. It also joined a community where aviation organisations trade warnings under labelled handling rules, a step the Atlas places in the first month of a new programme.
What readers can take from it
Students sometimes see operations and security as background noise to the real business of learning to fly. The story suggests otherwise. When a booking system fails on the day of a check ride, the cost lands on a student who has been saving for months. Anyone interested in how training costs add up may find it worth reading our pieces on trimming flight training expenses and the way aircraft choice shifts training cost, because avoidable disruptions are part of that picture too.
The broader lessons are plain. Do not let an automated rule act without evidence. Make every automated action easy to undo. Count outcomes, not volumes. Make sure a hunt ends in a record, and treat missing data as a finding in its own right. Habits like these are cheap to adopt, and they separate a routine that actually shields the academy from one that merely looks busy.

