Access review is the control sampled almost every time in ISMS and ITGC audits, and it’s where findings most often come up. The reason usually isn’t that the review was never done. It was done, but the auditor can’t follow its trail: where did the list come from, who looked at it, was the account marked “remove” actually closed? Below we explain how to build the chain from the source report to action closure in a form the auditor can retrace.
Why do access reviews often lead to findings?
In ISO/IEC 27001:2022 Annex A, A.5.18 (Access rights) addresses reviewing access rights at regular intervals, A.8.2 (Privileged access rights) restricting and managing privileged rights, and A.5.3 segregation of duties. On the ITGC side, the same topic is tested under access controls; for the overall framework, see the ITGC controls guide. The auditor looks at two things: the design of the review (scope, frequency, owner) and whether it was actually operated within the period. The second is harder.
The picture that turns into a finding in the field is almost always the same:
- The list reviewed is an Excel file whose source and extraction date are unknown.
- The IT manager approved their own administrator account.
- An account marked “remove” on the form is still active months later.
- Only Active Directory is in scope; the ERP, SaaS applications and network devices have been forgotten.
- Service and shared accounts have been dropped from the list on the grounds that they “don’t belong to a person”.
Scope: which systems, which accounts?
Derive the scope from the critical systems in the asset inventory and write it into the access control policy. How to build this part of the policy is explained in detail in the access control policy writing guide. The table below is an example framework for a mid-sized organization; the frequencies are examples and are set according to your own risk assessment.
| System | Source report | Reviewer | Example frequency |
|---|---|---|---|
| Active Directory / Entra ID | Export of users and group memberships | Unit managers; the ISMS owner for administrator groups | 6 months; privileged groups 3 months |
| ERP | User–role assignment report | Process owner (finance, purchasing, production) | 6 months |
| SaaS (CRM, M365 admin roles) | User/role export from the application | Application owner | 6 months |
| Database and server local administrators | Local group and DB role list | An approver other than the IT manager | 3 months |
| Firewall, switch, hypervisor | Local user and AAA role list | Infrastructure owner + ISMS | 3 months |
Write down the account types one by one too: personal accounts, privileged administrator accounts, service accounts, shared or common accounts, guest and supplier accounts, and accounts disabled but not deleted. The last one looks innocent. It isn’t. A disabled account keeps its group memberships; when someone re-enables it, all the old permissions come back.
Integrity and dating of the source report
The auditor has to trust the list the review is based on. Showing that a report produced by the organization is complete and accurate is the subject of the “information produced by the entity” (IPE) test in audit literature. In the field it’s simple: which system the report came from, with which query, by whom and when — all must be recorded. And once it goes to the reviewer, nobody touches the file.
An example of a reproducible export for Active Directory (requires the ActiveDirectory PowerShell module):
Get-ADUser -Filter * -Properties Enabled,LastLogonDate,whenCreated,Department,Manager |
Select-Object SamAccountName,Name,Enabled,LastLogonDate,whenCreated,Department,Manager |
Export-Csv -Path .\AD_Users_2026-Q3.csv -NoTypeInformation -Encoding UTF8
Get-ADGroupMember -Identity "Domain Admins" -Recursive |
Select-Object SamAccountName,objectClass |
Export-Csv -Path .\AD_DomainAdmins_2026-Q3.csv -NoTypeInformation -Encoding UTF8
Get-FileHash -Path .\AD_*_2026-Q3.csv -Algorithm SHA256 |
Export-Csv -Path .\AD_hash_2026-Q3.csv -NoTypeInformation
Record the command text, the run date and the hash values; keep the hash file in a location separate from the evidence folder or in the related ticket. That way the auditor can reproduce the report and see that it hasn’t changed since. There’s a trap: LastLogonDate comes from the lastLogonTimestamp attribute replicated between domain controllers, and because of the default update interval it can lag about 9–14 days behind the actual last sign-in. If you tie the “unused account” decision to an exact date, note this limit on the form.
Reconcile for completeness. Compare the total number of users in the report with the number in the system, and the active user list with HR’s active staff list of the same date. Every active personal account with no match in the HR list is tied either to an explanation (supplier, intern) or to a removal action. There’s no third option.
Who should the reviewer be? Segregation of duties
The reviewer is the person who knows whether the user needs that permission for their work. This is usually not IT but the unit manager or application owner. IT produces the report, makes technical role names understandable and applies the decisions. The rules directly questioned in audits are:
- Nobody approves the permissions of their own account; a senior manager or the ISMS owner looks at the IT manager’s privileged account.
- The person who assigns a permission and the person who reviews it aren’t the same; if this isn’t possible in a small team, an independent second check is defined.
- Conflicting role combinations in the ERP (for example creating supplier records and approving payments) are flagged in a separate column on the form.
Evidencing decisions and removed permissions
Every line on the form must have a decision: keep, remove or change. “Remove” and “change” are tied to a change or service request number. Hundreds of lines marked “OK” in bulk is the first thing that catches the auditor’s eye. A review that never produces a removal decision raises the same question.
An evidence package that causes no trouble in an audit is arranged in this order:
- Scope and period definition: which systems, which date, which policy clause.
- Source report, report query and hash record.
- The review form with a decision and justification on each line.
- The reviewer’s approval record (workflow approval or dated email).
- Request records and closure dates for removals and changes.
- A verification report pulled again after the actions, showing the removed permissions no longer exist.
The sixth step is the one most often skipped, and it’s the real evidence that the control works. A removal request may have been opened; the account may still be there. The Access Reviews feature in Microsoft Entra ID Governance records the decision and the reviewer in the system and can apply the result automatically; verify the licensing requirement in current Microsoft documentation. You can define this package’s place in the evidence matrix with the mapping method in the evidence matrix guide.
Additional steps for privileged and service accounts
Privileged accounts are reviewed more often, and the review looks not only at the account’s existence but also at how it’s used. In Active Directory, review the Domain Admins, Enterprise Admins, Schema Admins and Administrators groups with their nested memberships (-Recursive). This is a point that can be missed: an account that isn’t a direct member but gets administrator rights through a nested group. It’s also the first place to look in the console. In the ERP, profiles that give full rights (for example SAP_ALL in SAP), and the log review of emergency access sessions if they were used, are documented separately. For the technical hardening side, the article Active Directory security controls is a complement.
For service accounts, add these columns to the form: responsible person, application using it, interactive sign-in permission, password last changed date (PasswordLastSet) and the groups it belongs to. A service account with no owner is a finding on its own. If nobody owns it, nobody can close it either. Where possible, move to group managed service accounts (gMSA), whose passwords are managed by the system; the password age question disappears from the form entirely.
Stopping the review from becoming a formality
A review that sinks to “sign and move on” every period works on paper but doesn’t reduce risk. The signs below can be used by the ISMS team for its own quality check at the end of the period:
| Symptom | Mitigation |
|---|---|
| Hundreds of lines approved within minutes | Track approval time; split the list by the manager’s team |
| No removal decisions in any period | Have the ISMS team independently reassess a small sample |
| The manager doesn’t understand role names | Add a business-language description next to the technical role name in the report |
| The same kinds of errors come up every period | Look for the root cause in the joining, role change and leaving process |
For independent reassessment, set the sample size and selection method according to the principles in the sample selection in internal audit guide. If you want this cycle tested from an independent outside view, our access rights and user account audit and ITGC audit readiness services look at exactly this area.
Checklist
- In-scope systems and account types are written into the policy.
- Frequency and reviewer role are defined for each system.
- The source report query, extraction date and hash value are kept.
- The report is reconciled with HR’s active staff list.
- Nobody approved their own permissions.
- Every line has a decision and justification; removals are tied to a request number.
- A post-action verification report was pulled.
- Privileged groups were reviewed with their nested memberships.
- The owner and purpose of service accounts are recorded.
- If any period is overdue or missing, the reason and the catch-up are recorded.
Frequently asked questions
How often should access reviews be done?
ISO 27001 doesn’t give a fixed frequency; A.5.18 requires review at regular intervals, and the interval is set by the risk assessment and the access control policy. Common practice is a shorter cycle (for example quarterly) for privileged accounts and critical systems, and a six-monthly or yearly cycle for the rest. What really gets questioned in an audit is whether you met the frequency you wrote in the policy every period.
Is a manager writing “approved” by email enough evidence?
On its own, usually not. It must be clear which report the email approved, on what date and with what scope; there must be a keep-or-remove decision for each line, and a separate record must show the removed permissions were actually removed. An email carrying the decision form and the source report as attachments works as an approval record.
If a review finds an active account of someone who has left, is that a nonconformity?
The review catching the account shows the control works; the account staying open since the person left shows a gap in the leaving process. The auditor asks whether the account was used after the person left. Review the last sign-in information and related logs, record the result, and treat the leaving process as a corrective action.
Is the output of tools like Microsoft Entra Access Reviews accepted as evidence?
Yes. Records produced by the system that store the reviewer’s decision with a timestamp are strong evidence. Still, you need to address systems the tool doesn’t cover (ERP, local administrator accounts, network devices) separately, and verify that decisions were applied in the target system.