Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Security controls are classified by what they do in the timeline of an attack: preventive controls stop it, detective controls reveal it, and corrective controls restore what it damaged.

What Are Preventive, Detective and Corrective Controls?
Security controls are usually sorted two ways at once: by what they do in the lifecycle of an attack, and by how they are implemented. The first axis gives preventive, detective, and corrective controls. The second gives administrative, technical, and physical. Any given control has a position on both axes, and a door lock, a firewall rule, and a joiners-and-leavers process can all be preventive while being nothing like each other.
The classification earns its keep when you use it to find gaps. Lay your controls against both axes and the holes are usually obvious: heavy prevention, thin detection, and a corrective capability nobody has tested.
Preventive controls act before the event and aim to stop it happening. Multi-factor authentication, allowlisting, input validation, network segmentation, encryption at rest, background screening, and separation of duties all belong here. They are the most satisfying to buy and the most likely to be over-trusted, because every preventive control eventually meets an attacker who has found a path around it.
Detective controls act during or after the event and reveal that something has happened. Log collection and SIEM correlation, endpoint detection, file integrity monitoring, intrusion detection, access reviews, and internal audit are detective. Their value depends entirely on someone responding to the output, which is why an unmonitored alert queue is not a control at all.
Corrective controls act after the event to limit damage and restore service. Backups and restoration, incident response procedures, patching after a vulnerability is confirmed, account disablement, failover to a second site, and disciplinary processes all qualify. Recovery objectives are the measure that tells you whether these controls are adequate.
Most frameworks add a couple more. Deterrent controls discourage an attempt rather than blocking it: visible CCTV, warning banners, published audit activity. Compensating controls stand in for a required control the organisation genuinely cannot implement, providing a comparable level of protection by another route. PCI DSS formalises this, requiring documented justification and an assessor’s agreement before a compensating control is accepted, and version 4.0 added a customised approach for organisations meeting an objective in a different way.
Some taxonomies also list directive controls, meaning policies and instructions that tell people what to do. They matter in audit because a technical control with no governing policy is difficult to sustain after its author leaves.
| Administrative | Technical | Physical | |
|---|---|---|---|
| Preventive | Security policy, pre-employment screening, change approval | MFA, allowlisting, firewall rules, encryption | Door locks, turnstiles, cable locks |
| Detective | Internal audit, access recertification | SIEM correlation, EDR, integrity monitoring | CCTV, alarm systems, visitor logs |
| Corrective | Incident response plan, lessons learned | Backup restoration, automated isolation, patching | Fire suppression, replacement hardware |
Reading across a row shows whether a capability depends on a single implementation. Reading down a column shows whether a technology layer is doing all the work.
Budgets flow towards prevention because prevention is sold as a product and detection is sold as a programme. The consequence appears in every major incident report: attackers were present for weeks, the telemetry existed, and nobody was looking at it. Dwell time statistics have improved over the past decade, but intrusions discovered by a third party rather than by the victim remain stubbornly common.
The practical correction is to map your detective controls against attacker behaviour rather than against assets. Frameworks such as MITRE ATT&CK let you state honestly which techniques would generate an alert, which would merely be logged, and which would pass unseen. That exercise usually reallocates budget faster than any argument about products.
ISO/IEC 27001:2022 tags every Annex A control with attributes including control type, marked preventive, detective, or corrective, specifically so an organisation can analyse its selection this way. NIST SP 800-53 achieves something similar through its families, and the NIST Cybersecurity Framework functions line up closely: Protect is preventive, Detect is detective, and Respond and Recover are corrective.
The point of all of them is the same. Defence in depth is not a stack of preventive controls; it is coverage across the timeline, so that failure at one stage is caught at the next.
If you want evidence of which of your controls actually fire, our cyber exercise services test prevention, detection, and recovery together rather than in isolation.
Yes, and many do. Logging is detective in normal operation but supports corrective action by providing the evidence investigators need. Visible CCTV is both deterrent and detective. The classification is an analytical tool for finding gaps rather than a taxonomy that must be applied rigidly, so recording a control under two headings is perfectly reasonable.
A control implemented in place of a required one that the organisation cannot meet, providing a comparable level of protection by a different route. PCI DSS requires compensating controls to be documented with the constraint that prevents the original requirement, the alternative measure, and an assessor’s agreement. They are intended as a temporary measure, not a permanent substitute, and should carry a review date.
Preventive controls covering the common attack paths come first, because they remove whole classes of incident: multi-factor authentication, patching of internet-facing systems, and least privilege. Once those are in place, detection usually delivers more than additional prevention, since it determines whether a control failure becomes a contained event or a breach discovered by someone else.
The 2022 edition attaches attributes to each Annex A control, including control type with values of preventive, detective, and corrective, alongside information security properties, cybersecurity concepts, operational capabilities, and security domains. These attributes let organisations filter and analyse the control set in different ways, for example to check whether their selection is unbalanced towards prevention.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.