This is a short, practical primer for anyone who has been told to “look into SOC Ops” and does not know where to start. It assumes nothing. When you finish you will be able to follow a conversation about a security operations centre, spot the weak claims in a sales pitch, and know which words are worth learning first. Readers from aviation safety management will find several ideas familiar: reporting, risk ranking, investigation, recorded decisions and lessons carried forward.
Start here: a single paragraph
A security operations centre, or SOC, is the team and set of routines that watch an organisation’s computer systems for signs of misuse and decide what to do. SOC Ops is shorthand for running that function so it behaves like a dependable service: alerts arrive, people judge them quickly, outcomes get written down, and anything serious goes to the right person. The shorthand comes from the Ops Atlas, whose page on SOC Ops in the xOps School reference says the work itself has existed for decades while the exact name varies between organisations.
Mini-glossary: terms in the order they tend to appear
Keep this list handy. Each entry is deliberately short.
- Log: a dated record produced by a system, such as a sign-in or a file transfer.
- Alert: a message raised when a log pattern looks unusual enough to deserve a human glance.
- SIEM: software that gathers logs from many places, lets people search them and raises alerts. Think of it as the central filing cabinet with a very fast index.
- SOAR: software that automates repeatable steps, such as adding background details to an alert. The Atlas relays an Australian government warning that it usually does not suit immature setups, so a team should get its SIEM working first.
- Triage: the first quick look that sorts an alert into real, ordinary or unclear. The word is borrowed from emergency medicine; here it concerns systems, not people.
- Escalation: passing a case to a more experienced person, ideally with the evidence attached.
- Playbook: a written procedure for a recurring situation. Aviation people would call it a checklist.
- Disposition: the recorded ending of an alert. The Atlas names four outcomes: a true positive (a genuine threat), a benign positive (activity that happened but is harmless), a false positive (the alert was wrong) and an undetermined case.
- Pass-down log: the written note exchanged when one shift hands over to the next.
- Dwell time: the span during which an intruder goes unnoticed. A 2026 industry report quoted by the Atlas puts the worldwide median at two weeks.
- Median and 90th percentile: the middle value and the value that nine in ten cases beat. Used together they describe waiting times honestly, which a single average does not.
- MTTR: a vague abbreviation. It can mean respond, recover, repair or resolve, so always say which interval you mean.
- SOC-CMM: an open self-assessment for maturity, covering business, people, process, technology and services.
Things people confuse with a SOC
Beginners often mix up neighbours. A network operations centre keeps systems running; a SOC watches for misuse. “SOC 2” is an audit report from the AICPA about a company’s controls, and it has no direct link to the team. Physical security monitoring, such as cameras at a hangar door, is also a separate matter. If someone uses these terms interchangeably, politely ask which one they mean.
How a day actually goes
- Collect. Systems send logs to the central tool.
- Alert. Rules raise warnings. A good setup adds context, such as who the user is and how important the device is, before a person looks.
- Triage. An analyst checks and chooses a disposition.
- Escalate or close. Real or unclear cases move up; ordinary ones are closed with a reason.
- Hand over. The outgoing and incoming staff sign the pass-down log.
- Review. Weekly numbers show where the noise is and whether nights are covered.
The three numbers behind every queue
You do not need statistics to grasp the basics. How quickly alerts arrive, how long each takes to examine and how many analysts are present decide how long anything waits. The Atlas also reports a staffing rule of thumb from MITRE: keeping one seat occupied day and night calls for about five full-time people, and nobody should be alone on shift. That is why claims of “24×7 coverage” from a three-person team should prompt questions.
Who does what
Many centres use three levels. The first triages. The second investigates. The third brings specialists for hard problems. The Atlas says the first level should be somewhere people progress from, not a basement, and adds that tiers are optional; MITRE warns against rigid per-alert time limits. In small organisations, the same two or three people may play all roles, and the European cyber agency observes that in such places “CSIRT” and “SOC” are often synonyms.
Practical rules to remember
- Every alert ends with a recorded reason.
- Automation may isolate a staff device only when the rule is proven and the pace is limited; critical assets remain a human decision.
- Silencing a rule requires a responsible person and a lapse date.
- Outsourcing the watching does not outsource accountability.
- Judge providers on the quality of decisions, not on how quickly tickets close.
One quiet alert, start to finish
Imagine a flight school’s scheduling tool reports a sign-in from a city the instructor has never visited. The first-line analyst opens the matching playbook, asks whether the instructor is travelling, and finds a calendar entry that explains it. The disposition is “benign positive”, with a one-line note, and the alert is closed. Had there been no explanation, the case would have gone up a level with the sign-in record, the device details and the calendar check attached. In both outcomes someone can later read what happened and why, which is the real product of the whole system.
Where to go next
Pick one habit and try it for a month. The easiest is the four-way disposition with a one-line reason, because it costs almost nothing and immediately shows how much of your noise is ordinary behaviour needing context. After that, add a handover note, then a weekly look at medians and 90th percentiles split by shift. Only then think about buying tools. A team that can explain its own queue is better placed to choose software than one that starts with a demo.
