How to Write an Access Control Policy: A Section-by-Section Guide

An access control policy that meets Annex A 5.15–5.18 and 8.2: principles, identity lifecycle, privileged access, review and sample clauses.
Technical summary
Technology
ISO/IEC 27001:2022 Annex A 5.3, 5.15–5.18, 8.2, 8.5; ISO/IEC 27002:2022
Topic
Structure and clause writing for a topic-specific access control policy
Level
Intermediate
Prerequisites
  • An approved top-level information security policy
  • An information classification scheme
  • Application and system owners identified
In this article9 sections

If a leaver’s account is still open, or a user who has accumulated permissions over the years can access every share, the problem is usually not technical but in the policy. The access control policy is the most referenced topic-specific policy in an ISMS: account creation, administrator rights, password rules, permission reviews and leaving processes all rest on this text. Even so, in many organizations the policy is either a three-sentence statement of intent or has turned into a list of Group Policy settings. Neither works. Below we give the sections of an access control policy that meets the ISO/IEC 27001:2022 Annex A controls, what to write in each section and what not to write, with sample clauses.

Controls the policy must meet

The access control policy provides the basis not for a single control but for a group of interconnected controls. Before you start writing, mark on the SoA which controls this document will meet; you prevent scope creep from the start.

Check Topic Where it appears in the policy
5.15 Access control Principles, rules based on business and security requirements
5.16 Identity management The full lifecycle of identities, the shared account rule
5.17 Authentication information Allocation and protection of passwords and authentication information
5.18 Access rights Granting, changing, reviewing, removing permissions
8.2 Privileged access rights Restricting and managing administrator rights
5.3 Segregation of duties Conflicting duties not combined in one person
8.5 Secure authentication Situations that require multi-factor authentication

In its 5.15 guidance, ISO/IEC 27002:2022 recommends that access control rules be defined by a topic-specific policy and that the rules be based on business and information security requirements. For the relationship between the top-level policy and this topic-specific policy, see the information security policy writing guide.

Core principles: least privilege, need-to-know, segregation of duties

The first section sets the principles by which all later clauses will be interpreted. Write it short, but binding. Softeners like “as far as possible” open the principle to debate and make it harder to apply when exceptions are requested.

  • Need-to-know: A user accesses only the information they need to do their job.
  • Need-to-use and least privilege: A user is given the lowest level of permission needed for their job, only for as long as needed.
  • Default deny: Access that isn’t explicitly permitted is forbidden. When a new system goes live, access starts closed, not open.
  • Segregation of duties: Requesting, approving and implementing roles are separated; conflicting duties aren’t combined in one person, and if combining them is unavoidable, a compensating control is defined.
  • Individual accounts: Every access must be traceable to a single person. Shared accounts are used only where technically necessary and with an approved exception.

The access model is also written in this section: does the organization grant permissions role-based (RBAC), or use attribute-based (ABAC) rules in certain systems? The role definitions themselves go not into the policy but into system-specific role matrices.

Identity lifecycle: joining, change, leaving

This section links HR processes and IT operations. Two vulnerabilities you may come across come from here: a leaver’s account left open, and permissions accumulating with role changes. Both stem from the notification from HR to IT being undefined. The policy sets the trigger, the owner and the time target for each event; it leaves the operational steps to the procedure.

  • Joining: An account isn’t opened without HR’s approved joiner notification. Initial permissions are limited to the role’s standard permission set; additional permissions are granted with a separate request and approval.
  • Role change: While the new role’s permissions are granted, the old role’s permissions are removed at the end of a handover period. The handover period is defined in the policy with an upper limit; a handover with no upper limit means permanent permissions.
  • Leaving: Accounts are disabled on the leaving date. For departures that end on bad terms, disabling happens at the same time as the notification.
  • Long-term leave and inactive accounts: Accounts not used for a set period are disabled automatically or by periodic check. The period is the organization’s decision and is in the technical standard.
  • External users: Supplier, consultant and guest accounts are tied to an internal sponsor and opened with an end date. A supplier account with no end date keeps working years after the project ends.

The privileged access section

Annex A 8.2 requires the allocation and use of privileged access rights to be restricted and managed. The scope isn’t just domain administrators; database administrators, cloud subscription owners, network device management accounts, and administrator roles in the backup console and security tools also come in here. The backup console is often forgotten. In ransomware attacks, it’s often the administrator account in that console that deletes the backups. The policy should contain these decisions:

  • Privileged permissions are used through an administrator account separate from the person’s everyday user account; email and general internet use aren’t done with the administrator account.
  • Where systems allow, privileged permissions are granted time-limited (just-in-time access), not permanently.
  • Multi-factor authentication is mandatory for privileged accounts.
  • Use of privileged accounts is logged, and the logs are kept in an environment the person using the account can’t delete from.
  • Emergency (“break-glass”) accounts are listed separately; their use is recorded as an incident and reviewed afterwards.
  • Privileged permissions are reviewed more often than standard user permissions.

The technical implementation varies by platform; for Microsoft environments, the article Entra ID PIM configuration explains the implementation side of time-limited administrator rights. The policy, meanwhile, sets the rule without naming products.

Periodic review and exceptions

Annex A 5.18 ties not only the allocation, change and removal of access rights but also their review to the access control rules; ISO/IEC 27002:2022 guidance recommends this review be done at regular intervals and on role changes. The policy sets who does the review and how often. The review is done not by IT but by the owner of the information or application. IT produces the list and applies the removals requested. The frequency is tied to risk: privileged accounts and systems containing sensitive data are reviewed more often, low-risk systems less often.

The typical picture in the field: the application owner approves the hundreds-of-lines list sent to them without reading it, and the review turns into a signature. To prevent this, produce the list in language the owner understands (the business role, not the technical group name). How to evidence the review in an audit is explained in the access review evidence guide. What the policy needs to do here is define the mandatory output: who reviewed which list on what date, and which changes they requested.

The exceptions section is kept short: when the policy can’t be followed, from whom an exception is requested, who approves it, its maximum duration and where it’s recorded. The design of the exception process is the subject of the article policy exception management; the access policy only refers to that process.

Separating policy, standard and procedure

The most common reason access control policies break down is technical parameters getting into the policy text. Making a policy revision that needs top management approval because password length changed is unnecessary overhead. In the end, the revision gets postponed. The text and the system drift apart. The breakdown below shows which detail belongs to which document within an information security policy package.

Topic Policy Technical standard Procedure / work instruction
Password Strong authentication mandatory Length, banned password list, lockout threshold Password reset steps
MFA Mandatory for remote and privileged access Permitted methods Enrollment and device change
Inactive accounts Unused accounts are closed Threshold in days Reporting and closure steps
Permission review By the owner, periodic according to risk System-by-system frequency table List production and approval flow
Administrator account Separate account, time-limited rights Naming convention, session restrictions Privilege elevation work instruction

For points to watch when writing procedures and work instructions, see the guide to writing procedures and work instructions.

Sample policy clauses

The clauses below are a starting draft; adapt role names and periods to the organization’s actual structure.

4.1 Access to information systems is not granted without an approved request based on a business need. The request is approved by the user’s manager and by the owner of the system or information being accessed.

4.2 Each user is assigned a personal, unique account. Shared accounts are used only where technically necessary, with an exception record approved by the information security owner.

4.3 All accounts of leavers are disabled on the leaving date notified by HR.

4.4 Privileged rights are used through an account separate from the user’s everyday account, with multi-factor authentication.

4.5 System and information owners review access rights on the systems they’re responsible for at the intervals set in the Access Control Standard, and record the result.

4.6 When access in breach of this policy is detected, the rights are removed immediately and the incident is recorded under the information security incident management process.

Pre-publication checklist

  • Each of the related controls in the SoA is met by at least one clause in the policy.
  • Technical parameters are in the referenced standard, not in the policy.
  • The owner of each clause is given by role name and is consistent with job descriptions.
  • HR leaving and role change processes are consistent with the triggers in the policy.
  • Exceptions and breaches are linked to the relevant processes.
  • The policy has been announced to in-scope employees and a record of the announcement is kept.

The fastest way to see whether the policy is really applied in practice is an independent review of existing accounts and permissions. A single report from the directory shows how many of the policy’s clauses have stayed on paper. We do this work as part of our access rights and user account audit. For the full text of the control guidance, ISO/IEC 27002:2022 should be the reference.

Frequently asked questions

Should the access control policy and the password policy be separate?

A separate password policy isn’t mandatory. The access control policy sets the principle of strong authentication; parameters such as length, lockout threshold and the banned password list can be kept in a technical standard that’s easier to update.

How often should access rights be reviewed?

ISO 27001 doesn’t give a specific frequency; ISO 27002 guidance recommends review at regular intervals. The frequency is set by risk: privileged accounts and systems containing sensitive data are reviewed more often, low-risk systems less often, and this frequency is written in the standard.

How is segregation of duties achieved in a small IT team?

If request, approval and implementation roles can’t be fully separated, compensating controls are defined: logging of operations, periodic review of the logs by another person or a manager, and a second approval for critical changes.

Is it right to use product names in the policy?

Usually not. The policy sets rules that stay valid even if the product changes. Product and configuration details go into technical standards and work instructions; that way a system change doesn’t require a policy revision.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Compliance

We work with you on compliance with ISO 27001, IATF 16949, TISAX and similar standards, covering risk assessment, control design, documentation and audit preparation.

Knowledge Center

Other articles