Aviation safety managers already live by a loop that information-security teams have only recently put a name to: report a hazard, assess it, assign an owner, mitigate, and check that the mitigation worked. VulnerabilityOps is the security-side label for running that loop on known software weaknesses. The label is young, and the Ops Atlas describes it as emerging rather than standardised, yet anyone who has attended a hazard review board will recognise its habits. Here are eight of them, each tested against the evidence the Atlas cites.
1. Rank by evidence of real use, not by how frightening a label sounds
Hazard registers fill up with items scored on theoretical severity. The Atlas argues for a different order in software: evidence of active exploitation outranks a probability estimate, which in turn outranks a severity grade. The severity scale’s own maintainers say base scores should not carry a risk decision alone. The measurable reason is rarity, and it deserves a sentence of context: if only a few flaws in a hundred are ever attacked in a month, the discipline’s job is to find those few early rather than to polish the rest. According to the Atlas, only a small slice of published flaws, in the range of one to three percent per month, ever see attacker activity, so a ranking that ignores observed use wastes most of its effort.
2. Treat unknowns pessimistically
Every safety assessor has met the phrase “no data available” and learned to read it as a warning, not a reassurance.
Safety cases assume the worst when data is missing, and the clock design described in the Atlas follows suit. In the US federal directive it summarises, an unknown impact is read as total, the Atlas’s own local policy reads any untagged asset as public, and when a flaw is already beyond its deadline on the day it is found, the new due date is three days out. The directive reaches only certain agencies, but the principle travels. A missing field should never lower urgency.
3. Give every hazard one accountable owner and one due date
A further rule from the Atlas keeps the queue honest: reviewers may raise an item to a higher tier whenever context warrants, but nobody lowers one without a recorded reason. Priority can only be argued upward by default, which mirrors how a cautious safety board treats a disputed risk rating.
Shared responsibility dissolves under pressure. The discipline’s queue model has a tier, an owner and a date for each item, and nothing else is allowed to sit in limbo. The practical test is blunt: if you asked who is due to close this item by what day, could anyone answer without looking it up? For an aviation safety office, the equivalent is an action tracker whose every line has a name in the owner column.
4. Verify the mitigation, not the paperwork
This is the habit that separates serious programmes from ceremonial ones. The Atlas holds that closure must be earned with evidence. It accepts a rescan run with working credentials, a passing settings test, a rebuilt build whose fingerprint differs, or a challenge to the finding that was reviewed and upheld. A closed ticket proves a status change. It is the same distinction an auditor draws between a corrective action that was recorded and one that was observed working.
A case from the Atlas shows why a fix must include the damage already done. A late-2023 flaw in a widely used network gateway allowed attackers to take over logged-in sessions. The vendor’s advice went beyond patching and told operators to end every active and persistent session, because a hijacked session survives an update. In safety terms, the corrective action had to remove the existing exposure, not just prevent new ones.
5. Watch the reporting channel itself
Think of the last time a safety bulletin reached you late. The delay was probably in the routing, not in the analysis.
Hazard reports that go nowhere are worse than none. The Atlas recalls a well-known breach in which a scan ran about a week after a framework advisory and still missed the exposed portal, while the alert had gone to an outdated recipient list. Any safety office that has discovered a bulletin routed to a retired mailbox knows the chill. Audit distribution lists on a calendar, and treat a bounced alert as an event worth its own report.
6. Measure the flow, and distrust the average age
Safety performance indicators can mislead when they track a single figure. The Atlas proposes saturation: the number of new items arriving in a tier divided by the number verifiably closed, with a target below one. It then demonstrates a trap using a team whose oldest items dropped from thirty to twelve weeks of age while the pile grew by twenty-five a week. Newcomers had diluted the older ones. Read alongside a worked month in which a lower tier saw 600 arrivals against 520 closures, the lesson is clear: age alone can hide a queue that is drifting out of control.
A companion figure is coverage. The Atlas supplies a September example: 1,820 of 1,900 assets were covered by credentialed scans, near 96 percent, and eighty hosts had dropped out after a key rotation. Every safety manager knows that an inspection which skipped a bay says nothing about that bay.
7. Handle sensitive systems with a lighter touch
In aviation, nobody runs an intrusive diagnostic on a system that is in a flight-critical state. For plant and control equipment the Atlas gives parallel advice: probing a fragile device can knock it over, so listen passively instead. It adds that the logins a scanner uses are privileged secrets, that the findings database doubles as a map for attackers and needs protection, and that the group owning a product should never be the one to certify that a flaw leaves that product untouched.
8. Learn from the incidents that already happened
Safety culture runs on reviewing events, and the Atlas draws on a few. After the Log4Shell episode of late 2021, an official review reported that a single cabinet department burned about 33,000 working hours, and that among organisations holding software bills of materials, none said they had used one to locate the affected library. The Cloud Security Alliance’s suggested bar is to find a dependency within minutes. Another case, the xz backdoor of 2024, is used as an example for criticality tagging: an internal build host, not on the exploited list, receives a sixty-day clock under the sample policy, so the tag that says it is an internal build host decides the outcome.
Bonus: a maturity ladder you can place yourself on
Before climbing anything, find your current rung honestly. Programmes tend to overrate themselves by a level because dashboards look tidy.
The Atlas describes five levels. The bottom rung is scanning and reacting. Next come owned queues, then a risk-based policy, then managed flow, and at the top an adaptive stage where remediation is expressed as code. Safety managers can map this at once onto the familiar progression from reactive to proactive to predictive.
How to use this list
The eight habits are not a sequence. They are a set of checks, and a team can run a quick audit by asking one question per habit and recording a yes or no. A programme answering no to more than half has a clear first quarter ahead of it, and the answers also make a useful baseline for the monthly review.
Pick the habit your team is weakest at and fix only that this quarter. If you want to read the underlying definitions and sourcing, the Ops Atlas page covering VulnerabilityOps is the reference behind every claim above. Start small: choose one system, name its owner, set the date, and agree in advance what evidence will count as closed. The habits compound once the first loop completes cleanly.
