Security Operations in Practice
Follow a suspicious alert through the monitoring workflow: gather context, set its priority, escalate it to the right person and document each step.
One alert, from the first ping to a closed case
At 11:20 on a Thursday night, an alert appears on Kavya Rao's screen at Sentry Desk: Sameer Joshi's account has just downloaded 3,140 files from Tulsi Health's shared drive in twelve minutes. Sameer works in accounts and usually opens about forty files a day. Kavya does not panic and does not shrug. She follows a set of steps that every analyst at Sentry Desk follows, in the same order, for every alert. That repeatable path from an alert to a closed case is the monitoring workflow, and following one real alert through it shows what security operations look like in practice: checking facts, deciding how urgent something is, calling the right person and writing everything down.
The workflow every alert follows
Think of the 108 emergency ambulance service. A call-taker answers, asks a fixed set of questions, decides how urgent the case is, sends the nearest ambulance or passes the call to a doctor, and records every detail with the time. Nobody improvises the order of those steps in the middle of an emergency. A SOC works the same way. Systems are monitored all the time; a rule detects something and raises an alert; an analyst triages it, the quick first sort; doubtful alerts are investigated; real problems get a response or are escalated; everything is documented; and what was learned is used to improve the rules.
The loop closes at the end: every case teaches something that makes the next alert faster to judge or the next attack easier to catch.
The written steps for a common type of alert are called a playbook. Sentry Desk has a playbook for "large download", another for "sign-in from a new country", and so on. A playbook does not do the thinking, but it means a new analyst at 3 am asks the same important questions that a senior one would.
Investigating: turning an alert into a story
An alert is a single fact. Investigating a suspicious alert means collecting enough context around that fact to judge what really happened. Kavya works through a few plain questions. Who is the account, and what is their job? What exactly happened, and how much? When did it happen, and is that a normal time for this person? Where did it come from, which device and which location? And how different is this from the person's usual pattern, which analysts call the baseline?
Each answer comes from a different place. The SIEM, the system that gathers every log in one searchable place, shows the sign-in record. Sentry Desk's past cases show whether Sameer has done anything like this before. Tulsi Health's equipment list shows which laptops belong to staff. In four minutes Kavya knows that the download is about 80 times Sameer's normal amount, that it started at 11:05 pm from a laptop never seen on the account before, and that the two-step sign-in code was approved on Sameer's phone. That last fact could mean Sameer did it himself, or it could mean someone with his password sent a prompt and he tapped "Approve" half asleep.
One rule matters during an investigation: if you need to ask the user, contact them through a channel you already trust, such as a phone number saved in the staff list, never by replying to the account that may be in an attacker's hands. A message to a stolen mailbox reaches the attacker, not the employee.
Put Kavya's steps in the order a careful analyst follows them below. Each step produces something the next one needs, so think about what you must know before you can decide.
How urgent is it?
When an investigation shows that harm is happening or very likely, the event becomes an incident: something that actually damages or threatens the confidentiality, integrity or availability of information, that is, keeping it secret, keeping it correct and keeping it usable. Incidents do not all deserve the same speed. Incident prioritisation ranks them using two questions. Impact asks how much harm: how many people, how sensitive the data, which systems. Urgency asks whether it is still going on or spreading. Sentry Desk's contract with Tulsi Health uses four levels.
Priority can change as facts arrive. Kavya's case started as P2 and became P1 the moment Neha learned Sameer had not made the download.
A priority is never final. Kavya rated her case P2 because the facts were still uncertain; when they became clear, it went up. The opposite happens too, when a scary alert turns out to have an innocent explanation and drops to P4.
Escalating to the people who can act
Escalation means handing a case to someone with the authority or skill to do what the analyst cannot. Kavya can pause an account if Tulsi Health has approved that, but she cannot decide whether to tell customers, call the police or shut the shared drive for a day. Those decisions belong to Tulsi Health. An escalation path written in advance says who is called for what, in which order, and who to try next if the first person does not answer. For Tulsi Health it runs from Sentry Desk to Neha as the on-call contact, from Neha to Vikram for business and legal decisions, and to Rathi Softech when MedLedger is involved.
Each call carries facts and a clear request. The response team, Neha and outside specialists, now owns the case; Kavya's notes are their starting point.
A good escalation message says what happened, when, what is affected, what has been done so far and what decision is needed. It errs on the side of calling for anything that touches customer data, because in India many kinds of cyber incidents must be reported to CERT-In, the national computer emergency team, within six hours of being noticed, and that clock does not wait for a convenient morning. Escalating everything wears people out; escalating too little leaves the decision-makers learning about an incident from a customer.
The four situations below come from one busy Monday at Tulsi Health. Decide which case comes first, who must hear about it and what the handover note should contain.
Writing it all down
Security operations documentation is the written record of each case: the alert, every check, every decision and who made it, each with a time. Kavya's case ticket is a real example of what that looks like. It matters for four reasons. The next shift can pick up the case without repeating work. The response team starts from facts rather than memory. If customers, regulators or the police ask questions later, the notes show what was known and when. And reading old cases is how the SOC improves its rules.
Every note has a time, one fact or action, and who did it. Anyone reading it next can tell exactly what was known at 23:33 and what changed at 23:52.
Good notes record facts, not feelings: "3,140 files in 12 minutes" rather than "huge suspicious download". They say where each fact came from, so it can be checked. And they are written as the work happens, not reconstructed the next morning, because memory is the first thing to blur under pressure. When Kavya's case closes, a short review asks one more question: what would have caught it sooner? Here the answer was a new SIEM rule linking an approved two-step prompt late at night with a large download, which now runs for every Tulsi Health account.