bastion

Understanding Security Events and Incidents

Build this skill

Separate ordinary security events from suspicious activity and confirmed incidents, recognise indicators of compromise, and assign a severity that sets how fast to respond.

Forty thousand things happen before lunch

By lunchtime on an ordinary Tuesday, Tulsi Health's computers have noted more than 40,000 small happenings. Farida logs in to MedLedger at the Kothrud counter, a customer's phone joins the guest Wi-Fi in Nashik, Rohan's app update goes live, and someone types a password into Sameer's email account. Each of these is a security event: anything that happens in a computer system that could matter for security, such as a login, a file being opened or a setting being changed. Almost every event is harmless. Neha Kulkarni's IT and security team has two people, so their real skill is not looking at everything but spotting the few events that mean trouble, and deciding how fast to act on them.

Neha's screen · events so far today 41,236 Time Who What Where 09:02farida.sEmail login OKKothrud 09:03KOT-1MedLedger startedKothrud 09:05guestJoined guest Wi-FiNashik Rd 09:07rohan.iApp update liveOffice 09:09sameer.j6 failed loginsBhopal 09:10sameer.jEmail login OKBhopal 09:11sameer.jMail now forwarded outBhopal 09:12arjun.pInvoice printedNashik Rd Every line is an event 41,236 today, almost all of them ordinary work ! Suspicious activity 6 failures, then a success from a city nobody works in ✕ Incident Sameer's mail is now being copied to an outsider

Most lines are ordinary work. Two lines look odd and deserve a second look, and one shows harm actually happening.

Events: everything the systems notice

A police station gives a good picture of how this works. Its daily diary records everything that comes through the door: a lost wallet, a noise complaint, a visitor asking for directions. Only a few entries ever turn into a case that officers investigate. Computer systems keep a similar diary of events, and most entries are as dull as "Farida logged in at 9:02 am".

Events matter even when they are dull, because they are the raw material for everything else. A single failed login means nothing; someone mistyped a password. Twelve failed logins on one account in two minutes, followed by a success from a city where no Tulsi Health employee works, is a story. Defenders can only notice that story because each small event was recorded.

Eventsabout 40,000 a daySuspiciousabout 20 a daystands outCheckedNeha looks closerinvestigateIncidentsa few a monthconfirmed

Each stage keeps far fewer items than the one before. The work is narrowing the flood down to the few things that really need a person.

Suspicious activity: events that deserve a second look

Suspicious activity is an event, or a pattern of events, that does not fit normal behaviour and could be part of an attack, but might still have an innocent explanation. Common examples are many failed logins on one account, a login at 3 am by someone who works days, a login from a new city, a staff member opening files they have never needed, or someone downloading far more data than usual. The key word is "could". Vikram might be logging in from Mumbai because he is at a supplier meeting there, and Riya might be opening the contracts folder because Sameer asked her to.

So suspicious activity is a question, not an answer. Neha's job is to check it: ask the person, compare with what they normally do, look at what happened next. The question is usually settled in minutes, one way or the other.

Indicators of compromise: clues that an attack already happened

After a burglary, police look for clues: a broken latch, muddy footprints on the stairs, a cupboard left open. Computer systems leave similar clues. An indicator of compromise, or IoC, is a piece of evidence that a system has probably already been broken into, "compromised" being the security word for taken over or tampered with. Where suspicious activity says "this looks odd", an IoC says "this is the kind of trace attackers leave".

Known-bad addressA billing laptop contacts aweb address that appears ona list of criminal serversMalware fingerprintA file on the shared drivematches the fingerprint of aknown ransomware programHidden mail ruleSameer's mailbox has a new rulequietly forwarding everyemail to an outside addressUnknown programA program nobody installedstarts every time theKothrud laptop is switched on

Each clue is specific enough to check. Security companies share lists of known-bad addresses and file fingerprints, so one victim's clues help everyone else look.

A few terms in those cards need plain words. A web address is the name you type to reach a website, and behind it sits an IP address, a number that identifies a computer on the internet, rather like a postal address. A file's fingerprint, also called its hash, is a short code calculated from its contents; two identical files always have the same fingerprint, so a file can be recognised as known malware without opening it. IoCs are valuable because they can be searched for. Once Neha knows the fingerprint of the ransomware that hit another pharmacy chain, she can check all 14 billing laptops for it in minutes.

When an event becomes an incident

A security incident is an event, or a chain of events, that actually harms, or seriously threatens, the confidentiality, integrity or availability of information or systems, or that breaks the organisation's security rules. Confidentiality means only the right people can see something; integrity means it has not been wrongly changed; availability means it works when needed. An incident is the station diary entry that becomes a case.

The line between "suspicious" and "incident" is crossed when checking confirms harm or a real threat. Twelve failed logins on Sameer's account are suspicious. When Sameer says he was asleep in Pune at the time, the Bhopal login was not him, and a new rule is forwarding his mail outside, it is an incident: someone else is inside his account and reading company email. From that moment the team stops asking "is something wrong?" and starts asking "how do we stop it and repair it?"

Sort the situations below into three buckets. Ask yourself two questions in order: does it fit normal work, and if not, has harm already been confirmed?

How serious is it? Incident severity

Not every incident needs the same speed. At a hospital emergency desk, a nurse sees a child with a mild fever and a man clutching his chest at the same moment. Both are patients, but the chest pain goes in first, because a delay there could cost a life. This sorting is called triage. Incident severity is the same idea for security: a rating of how much harm an incident is causing or could cause, which decides who is called and how quickly they must act.

Teams usually judge severity by asking four questions. What is affected: customer data and payments, or a single printer? How many people or systems: one laptop or all 14 stores? Is it still happening or spreading, or has it stopped? Can the business keep working? Most organisations turn the answers into three to five levels, each with a target response time.

Criticalcustomer data leaving nowact within minutesHighone account taken overwithin the hourMediumone laptop infectedsame dayLowunclicked phishing emailthis week

Severity sets the clock. A low incident still gets handled, just not at the cost of a critical one waiting.

Severity can change as the team learns more. An infected laptop rated medium becomes critical the moment Neha finds it has been sending customer records out for a week. That is why the rating is revisited at every step, rather than fixed at the start and forgotten.

Try the triage yourself below. You are on duty with Neha for one long day, and each situation needs a severity. Use the four questions: what is affected, how much, is it still happening, and can the stores keep working?

Follow your curiosity.

↑ ↓ to explore · Enter to open · Escape to close