Fundamentals

What is a security control?

Updated

A security control is a safeguard or countermeasure that an organization applies to a system to protect the confidentiality, integrity, and availability of that system and its information. Controls are the unit of work in every security framework, from SOC 2 to NIST SP 800-53, and they are what auditors test. This guide covers the types of controls, real examples, and how controls are tested and automated.

Key takeaways

  • A security control is a safeguard or countermeasure that protects the confidentiality, integrity, and availability of a system and its information (NIST SP 800-53 Rev. 5).
  • Controls are classified by function (preventive, detective, corrective) and by nature (administrative, technical, physical). Most real controls fit one of each.
  • A compensating control substitutes for a recommended control when the original cannot be implemented, as long as it provides comparable protection.
  • Testing a control means checking both design and operating effectiveness, using examine, interview, and test methods.
  • GRC engineers implement technical controls as code and test them continuously with tools like AWS Config, OPA, and CI pipeline checks.

What does "security control" mean?

NIST SP 800-53 Rev. 5 defines security controls as the safeguards or countermeasures prescribed for an information system or an organization to protect the confidentiality, integrity, and availability of the system and its information. That definition is broad on purpose. A control can be a line of Terraform, a door lock, a quarterly access review, or a written incident response procedure.

Three related terms are worth separating. A requirement is what a framework or regulation asks for. A policy is the organization's own statement of what it will do. A control is the specific mechanism that makes the policy true in practice. "Users must authenticate with MFA" is a policy statement. An identity provider configuration that rejects password-only logins is the control.

Every well-written control also has a control objective: the risk it exists to reduce. Account management (AC-2 in SP 800-53) exists so that only authorized people hold access. Keeping the objective in view is how you judge whether a control is working, and whether a different mechanism could meet the same objective. For the bigger picture of how controls fit into governance and risk, see what GRC is.

What are the types of security controls?

Controls are classified along two independent axes. The first describes what the control does in time: before, during, or after an event. The second describes how the control is implemented. A single control usually has one label from each axis, so MFA is both preventive and technical.

TypeWhat it doesExample
PreventiveStops an unwanted event from happeningMFA, least-privilege IAM roles, blocking public storage buckets
DetectiveIdentifies an event while or after it happensAudit logging, intrusion detection, configuration drift alerts
CorrectiveLimits damage and restores normal operationBackups and restore, automated remediation, incident response
AdministrativePolicies, procedures, and people-based processesSecurity training, access reviews, vendor due diligence
TechnicalImplemented in hardware, software, or firmwareEncryption at rest, firewalls, endpoint protection
PhysicalProtects facilities, equipment, and mediaBadge readers, cameras, locked server cages

The administrative, physical, and technical split is familiar from the HIPAA Security Rule, which organizes its requirements into those three safeguard categories. Some models add deterrent controls (warning banners) and recovery controls as separate functions.

A compensating control is a special case. NIST describes it as a control employed in place of a recommended control that provides equivalent or comparable protection. If a legacy application cannot support MFA, restricting it to an isolated network segment reachable only through an MFA-protected bastion can compensate. Auditors will expect a documented rationale for why the substitute meets the original objective.

What are examples of security controls?

The examples below use control identifiers from NIST SP 800-53 Rev. 5, the catalog most other frameworks map back to. The same safeguards appear, under different numbering, in SOC 2, ISO 27001 Annex A, and the NIST Cybersecurity Framework. For a deeper look at the catalog itself, read what NIST 800-53 is.

ControlSP 800-53FunctionNature
Multi-factor authentication for workforce usersIA-2PreventiveTechnical
Encryption of data at restSC-28PreventiveTechnical
Event logging for security-relevant actionsAU-2DetectiveTechnical
Continuous system monitoring and alertingSI-4DetectiveTechnical
Periodic review of user accountsAC-2DetectiveAdministrative
System backupsCP-9CorrectiveTechnical
Security literacy training and awarenessAT-2PreventiveAdministrative
Physical access control to facilitiesPE-3PreventivePhysical

How do you test a security control?

Control testing answers two questions. Design effectiveness asks whether the control, as described, would meet its objective. Operating effectiveness asks whether it actually ran as designed over a period of time. A SOC 2 Type 1 report covers design at a point in time; a Type 2 report adds operating effectiveness over a review period. The SOC 2 guide covers the difference in detail.

NIST SP 800-53A, the assessment companion to SP 800-53, describes three assessment methods that assessors use to gather evidence:

  • Examine: Review documents, configurations, logs, and other artifacts. Reading an IAM policy or a Terraform module is an examine step.
  • Interview: Talk with the people who own and operate the control to confirm how it works in practice.
  • Test: Exercise the mechanism directly and compare actual behavior to expected behavior, such as attempting a password-only login.

Traditional testing samples. An auditor picks 25 new hires and checks each one for a completed access request. Automated testing evaluates the whole population continuously, which is why engineered controls produce stronger evidence. What counts as sufficient evidence is covered in what audit evidence is.

How do GRC engineers implement security controls as code?

A GRC engineer treats a technical control as something to deploy, version, and test like any other piece of infrastructure. Take SC-28, protection of information at rest. Instead of a screenshot of a bucket settings page, the engineer deploys an AWS Config managed rule that evaluates every S3 bucket whenever its configuration changes:

Resources:
  S3EncryptionRule:
    Type: AWS::Config::ConfigRule
    Properties:
      ConfigRuleName: s3-bucket-server-side-encryption-enabled
      Description: "SC-28: S3 buckets must have default encryption enabled"
      Source:
        Owner: AWS
        SourceIdentifier: S3_BUCKET_SERVER_SIDE_ENCRYPTION_ENABLED
      Scope:
        ComplianceResourceTypes:
          - AWS::S3::Bucket

That rule is a detective control. The matching preventive control lives earlier in the lifecycle: a policy-as-code check in the CI pipeline that fails any Terraform plan creating an unencrypted bucket. Pairing both gives you defense in depth plus a continuous record of compliance status. The mechanics are covered in what policy as code is and continuous compliance monitoring.

Administrative and physical controls resist full automation, but engineering still helps. Access reviews can be generated from identity provider APIs, routed as tickets, and closed with a recorded decision. Every control you codify is one less control you re-prove by hand each audit cycle. When a control fails, the gap becomes a risk that belongs in a risk register.

Frequently asked questions

What is a security control?

A security control is a safeguard or countermeasure an organization puts in place to protect the confidentiality, integrity, and availability of a system and its information. NIST SP 800-53 Rev. 5 uses this definition. Examples include multi-factor authentication, audit logging, encrypted storage, backups, and security awareness training.

What are the three main types of security controls?

Controls are usually grouped two ways. By function, they are preventive (stop an event), detective (find an event), and corrective (fix or recover from an event). By nature, they are administrative (policies and procedures), technical (implemented in systems), and physical (protecting facilities and hardware).

What is the difference between a control and a policy?

A policy states what the organization requires, such as "all production data must be encrypted at rest." A control is the specific mechanism that meets the requirement, such as default encryption on every storage bucket enforced by infrastructure code. Auditors test controls to determine whether the policy is actually being followed.

What is a compensating control?

NIST defines a compensating control as a management, operational, or technical control used in place of a recommended control that provides equivalent or comparable protection. For example, if a legacy system cannot support MFA, an organization might restrict it to a segmented network reachable only through an MFA-protected bastion host.

How do auditors test a security control?

Auditors evaluate design (would the control meet its objective if it operated as described) and operating effectiveness (did it actually operate over the period). NIST SP 800-53A describes three assessment methods: examine documents and artifacts, interview people, and test the mechanism directly.

Can security controls be automated?

Many technical controls can be both enforced and tested automatically. A GRC engineer might deploy an AWS Config managed rule that checks every S3 bucket for default encryption and records compliance status continuously, which produces evidence of operating effectiveness without a manual sample.

Build controls you can prove

GRC Engineering Club members learn to design, deploy, and test controls as code in hands-on cloud labs, alongside a community of practitioners doing the same work. Membership is $70 a year.