bastion

Understanding Security Controls

Build this skill

Classify controls as preventive, detective or corrective and assemble layers so that no single failure exposes an asset.

A lock, a camera and a spare key

Anything you put in place to make a bad event less likely, or less harmful when it happens, is a safeguard. In cybersecurity a safeguard is called a security control. A control can be a piece of software, a setting, a rule that people follow, or even a locked cupboard. What makes it a control is that it reduces a specific risk to a specific asset.

Think about the housing society you live in. The gate has a lock and a watchman, which try to stop strangers from walking in. The corridor has a CCTV camera, which does not stop anyone but tells you afterwards that someone came. And the society keeps an insurance policy and a set of spare keys with the secretary, so that if something is stolen or a lock is broken, life gets back to normal quickly. Those three things do three different jobs, and every control in cybersecurity does one of the same three jobs.

1. Prevent Stop it happening Locked door Login code 4829 Enter Two-step login Blocks the attack 2. Detect Notice it fast REC CCTV ! New login Farida's mail, Bhopal, 2:14 am Login alert Raises the alarm 3. Correct Repair the damage Backup Tue 2 am Offline copy restore ✓ Restored Billing laptop Gets you working again

The three jobs a control can do. Prevention tries to stop the attack, detection raises the alarm when prevention misses, and correction gets you working again.

Neha Kulkarni runs IT and security for Tulsi Health with a team of just two people. She cannot buy every product on the market, so for each control she asks one question first: which of the three jobs does it do? The answer tells her what it will and will not save her from.

Controls that stop trouble before it starts

A preventive control tries to stop a bad event from happening at all. At Tulsi Health, examples include an email filter that blocks dangerous attachments, a locked cupboard for the shared-drive computer, and training that teaches staff to recognise fake messages. Another is two-step login, also called multi-factor authentication (MFA): after typing a password, you must also enter a one-time code sent to your phone. A thief who has only the password is stuck at the second step.

Prevention is the most attractive kind of control, because the cheapest incident is the one that never happens. Its weakness is that no preventive control is perfect. Farida Shaikh might type both her password and her one-time code into a very convincing fake page, or a brand-new piece of harmful software might slip past the email filter because nobody has seen it before. If prevention is the only thing you have, a single miss goes completely unnoticed.

Controls that notice when something gets through

A detective control does not stop anything. Its job is to notice that something has gone wrong and tell a person. Tulsi Health's examples are an SMS alert to Neha when someone logs in to a staff mailbox from an unusual place, the CCTV over each counter, antivirus warnings, and the monthly stock count that shows when medicines are missing. Many detective controls rely on logs, which are diaries that computers keep of who did what and when.

The society gate register is the everyday version. It does not stop a visitor, but if a bicycle goes missing, the register shows who came in that afternoon. A detective control is only as good as the person reading it, though. An alert that lands in an inbox nobody opens detects nothing, so every alert needs an owner and a plan for what happens next. Speed matters too, because the longer an intruder goes unnoticed, the more damage they can do.

Controls that repair the damage

A corrective control limits the harm after an incident and brings things back to normal. Restoring files from a backup, rebuilding a laptop from a clean copy, resetting stolen passwords, blocking a misused account and installing a fix for a software flaw are all corrective. So is a written incident response plan, a checklist that says who does what in the first hours of a crisis, so nobody wastes time deciding.

Corrective controls fail quietly if nobody tests them. Tulsi Health's nightly backup is copied to the same shared drive that the billing laptops use. If harmful software scrambles that drive, the backup is scrambled along with everything else. A copy kept offline, disconnected from the network, cannot be reached by an attack on the network, which is why Neha wants one even though the nightly backup already exists.

PreventiveJob: stop it happeningTwo-step loginEmail filterStaff trainingDetectiveJob: notice it fastLogin alertsCCTV and logsMonthly stock countCorrectiveJob: repair the damageRestore from backupReset stolen passwordsRebuild a laptop

Sort any control by asking what it does at the moment of an attack: stop it, spot it, or fix it.

Sorting controls by their job is a skill you will use constantly, because a list of controls that are all the same type has a gap you can see at once. Place each Tulsi Health control in the bucket for the job it does.

Why one strong wall is never enough

Picture an old hill fort in Maharashtra. Attackers first faced a steep climb, then a ditch, then thick outer walls, then a gate with iron spikes and guards on the towers who could see them coming and call for help. Even if they broke through, the treasury sat in an inner fort with walls of its own. No single defence had to be perfect, because each one slowed the attackers and gave the next one a chance.

Security people call this defence in depth: protecting an asset with several layers, so that when one layer fails another one still stands between the attacker and the asset. The important word is "different". Two email filters from two companies tend to miss the same clever email, so they count as nearly one layer. A filter plus trained staff plus two-step login plus an alert plus a backup are five different kinds of protection that fail for different reasons, so the chance of all five failing at once is much smaller.

Staff trainingSpot fake emails and callsPreventEmail filterBlocks bad attachmentsPreventTwo-step loginPassword alone is not enoughPreventEndpoint protectionStops harmful softwarePreventUnusual-activity alertsTell Neha within minutesDetectOffline backup copyRestores what was lostCorrectCustomer recordsThe asset being protectedAsset

An attack has to get through every layer above the asset. Even if all the preventive layers fail, the alert and the offline backup still limit the harm.

Walk one attack through the layers. A fake "your mailbox is full" email reaches Farida, and she does not spot it, so training has failed. The email filter does not recognise the link, so that layer has failed too. She enters her password on the fake page, but the attacker in another city still needs the one-time code on her phone, and the attack stops at the third layer. Had it got further, the alert would have told Neha, and the offline backup would have restored anything damaged.

Plan as if someone is already inside

The fort's builders assumed attackers might get past the outer wall one day. Modern defenders make the same assumption on purpose, and call it the assume breach principle: plan as though an attacker is already inside some part of your systems, because sooner or later one will be. A bank works this way. It does not trust its front door alone; the locker room has its own door, a register, a camera, and every locker needs two keys, the customer's and the bank's.

For Neha, assume breach turns into three plain questions. If Rohan Iyer's laptop were controlled by an attacker right now, what could it reach? How quickly would anyone notice? And how quickly could Tulsi Health recover? The answers push her towards giving each account only the access its job needs, keeping customer Wi-Fi apart from the billing laptops, setting up alerts that a person actually reads, and practising a restore from the offline copy.

Keep them outOne strong wall at the edgeEveryone inside is trustedMoney goes on preventionA breach is a surpriseAssume breachSeveral layers of different typesEach account gets only what it needsAlerts that a person acts onRestores are practised in advance

Assume-breach thinking does not give up on prevention. It adds the layers that matter once prevention has failed.

Now try building the layers yourself. Tulsi Health has a yearly budget of ₹1,50,000 for new controls, and six things are likely to go wrong this year. Switch controls on, press Run, and see which attacks are blocked, which are only detected or recovered, and which cause damage that nobody notices. Look for a combination that handles every attack within the budget.

Notice that some attacks have no preventive answer on the list at all. Nothing in the budget stops a forgotten account that still has a valid password, so the alert is the only layer that can catch it, and nothing prevents an honest mistake, so only a backup can undo it. Notice also that money spent on two similar preventive controls buys less than the same money spread across preventing, detecting and correcting. That is defence in depth and assume breach working together.

Follow your curiosity.

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