Skip to content
12 Best Practices for an Office 365 Security Audit episode artwork

Oct 9, 2026 <1 min

12 Best Practices for an Office 365 Security Audit

from PostSphere Hosted by Post Sphere

People still call it Office 365, even though Microsoft renamed the suite years ago. The name stuck because it describes what most organisations actually use day to day: email, Word, Excel, Teams, SharePoint and OneDrive. Behind those apps sits a tenant full of settings, and the job of an Office 365 security audit is to check whether those settings protect the business the way everyone assumes they do. An audit can go well or badly, and the difference is rarely about technical skill alone. Good audits start with a clear scope, ask for evidence instead of assurances, and end with a report that somebody can act on. Poor ones produce a long PDF that sits in a shared drive and gets quoted once during an insurance renewal. Whether you are running the review yourself or commissioning an Office 365 security audit from an outside firm, these 12 practices will help you get a result that is useful. Each one comes with a common pitfall, because knowing where audits go wrong is half the battle. 1. Start with the business question, not the settings list Before anyone opens a portal, write down why the audit is happening. Is it driven by a cyber insurance renewal? A customer contract that demands evidence of controls? A recent phishing incident? A planned Copilot rollout? The reason shapes everything that follows, including which workloads matter most and how deep to go. A law firm worried about client confidentiality will care most about sharing, guest access and mailbox permissions. A manufacturer with plant managers on shared devices will care more about Intune and sign-in controls. Common pitfall: treating the audit as a generic checklist exercise. A checklist finds generic problems. A question-led audit finds the problems that matter to you. 2. Define the scope in writing Scope covers four things: which tenants, which workloads, which evidence period and which audience for the report. Write them down and get sign-off from whoever is paying and whoever owns the systems. Be explicit about the boundaries. If you have a hybrid Active Directory environment, say whether on-premises identity is in or out. If you have a second tenant from an acquisition, name it. If third-party tools such as a mail filter or backup service sit around Microsoft 365, decide whether they are included. Common pitfall: a vague scope leads to arguments later. "We thought you were checking SharePoint" is an expensive sentence. 3. Use read-only access and agree how evidence is gathered An audit does not need the power to change your tenant. Use read-only roles such as Global Reader and Security Reader where possible, or work through supervised sessions where your administrator shares a screen and the auditor directs the review. Another common route is for your team to run approved exports and hand over the files. Document who authorised the access, which accounts were used, and when the access was removed. Do not share passwords, and do not create standing admin accounts for an outside party unless you absolutely must. Common pitfall: handing over a Global Administrator credential because it is quicker. It is quicker, until an audit account becomes a target or an auditor's laptop is lost. 4. Review identity before anything else Most attacks on Microsoft 365 begin with a stolen identity, so identity deserves the first and deepest look. That means MFA coverage by user population, Conditional Access policies and their exclusions, legacy authentication, privileged role assignments, emergency access accounts, guest users and service accounts. Work from the question "who is not covered?" instead of "is the policy on?" A policy can be enabled and still protect only 70 percent of your users. Common pitfall: checking whether MFA is "enabled" in the tenant and stopping there. Per-user MFA states, Conditional Access, Security Defaults and authentication methods all interact, and the effective result can differ from the setting names. 5. Baseline against recognised standards, then adjust for your business Pick a reference point so that findings are consistent and defensible. Common choices include the CIS Microsoft 365 Foundations Benchmark, Microsoft's own security baselines and recommendations, and the CISA SCuBA baselines for Microsoft 365, which were written for US federal agencies but are public and widely used elsewhere. A baseline is a starting line, not a verdict. Some benchmark items will not suit your business, and some will need licences you do not hold. Record each deviation with a reason. Common pitfall: chasing a perfect score against a benchmark. A deviation with an owner and a documented reason is a better outcome than a blind fix that breaks a business process. 6. Check mail flow and email protection as a system Email security is not one setting. It is a chain: DNS records for SPF, DKIM and DMARC; connectors; transport rules; anti-spam, anti-phishing and anti-malware policies; Safe Links and Safe Attachments if you have Defender for Office 365; mailbox auditing; and forwarding controls. Test the chain from both ends. Look at inbound protection and outbound controls, and check whether your most targeted users, such as executives and finance staff, sit inside stronger policies. Common pitfall: reviewing policies one at a time without asking how they combine. A strict policy applied to nobody, or overridden by a transport rule, protects no one. 7. Trace sharing back to owners and data External sharing is where good intentions create risk. Review tenant-level sharing settings, then site-level overrides, then actual sharing activity. Look at the number of anonymous links, the age of those links, guest accounts in Teams and SharePoint, and which sites hold sensitive content. Where you can, link a sharing configuration to a person. Every site, team and group should have an owner who can say why it is shared the way it is. If nobody can, that is a finding. Common pitfall: counting links without looking at the content behind them. A thousand links to a public brochure library are far less worrying than ten links to an HR folder. 8. Review role-based access control and privileged access Role-based access control only works if roles match real jobs. List every holder of privileged roles across Entra ID, Exchange, SharePoint, Teams and Intune. Ask whether each assignment is still needed, whether it should be permanent or time-limited, and whether the account is separate from daily-use accounts. Do not forget indirect privilege. An application registration with broad Graph permissions, or a group nested inside a role-assignable group, can act like an administrator without appearing in the admin list. Common pitfall: reviewing only the Global Administrator list. Other roles, such as Exchange Administrator or Privileged Authentication Administrator, can be just as damaging. 9. Look at third-party apps and OAuth consent Non-human access is a growing part of the attack surface. Review enterprise applications, app registrations and consent grants. Note which apps can read mail, files or directory data, who granted the consent, whether the publisher is verified and when the app last signed in. Also check how consent is governed. Can any user approve any app? Is there an admin consent workflow? Are risky permissions blocked by default? Common pitfall: forgetting that users, not just admins, can consent to apps in many default configurations. 10. Verify logging, retention and alerting You cannot investigate what you cannot see. Confirm that the unified audit log is enabled in Microsoft Purview, check which audit tier your licences provide and how long records are kept, and confirm mailbox auditing is on for the accounts that matter. Review whether any alert policies exist and who receives them. Run a test. Search for a known recent action, such as a file download or a sign-in, and see if the evidence appears. A setting that says "on" is less convincing than a search result that returns data. Common pitfall: mixing up a missing record with a missing control. If a log is not available, the finding is "unable to verify," not "control absent." Good reports keep those two statements apart. 11. Consider data protection, and be realistic about it Data loss prevention, sensitivity labels and retention policies in Microsoft Purview can reduce the damage of mistakes and leaks. Many tenants have licences for these features but have not configured them, or configured them in "audit only" mode and never moved on. An audit should note what exists, where it applies and how it is being used. It should also be honest about maturity. A small firm that has labelled its three most sensitive document libraries is in a better position than a larger firm with a grand classification scheme that nobody follows. Common pitfall: recommending a sweeping data classification project to an organisation that cannot staff it. Prioritise a small number of high-value locations first. 12. Write findings that someone can act on, and track them to closure The report is the product. Each finding should state what was observed, where the evidence is, why it matters, what to do about it, who should own it and how to confirm it is fixed. Prioritise by likelihood and impact, and mark dependencies, such as "requires Entra ID P1." Walk through the draft with your team before it is final. They will spot factual mistakes and business dependencies the auditor could not have known. After that, set dates, assign owners and schedule a revalidation so you know the fixes worked. Common pitfall: an audit that ends at the report. Without owners, deadlines and a follow-up check, the same findings reappear a year later. A sample audit timeline Timelines vary, but a focused Office 365 security audit often follows a pattern like this: Days 1 to 2: agree scope, confirm licences, arrange access, collect any existing policies and documents. Days 3 to 7: review identity, email, collaboration, devices and logging. Capture evidence as you go. Days 8 to 9: validate findings with your administrators, correct anything that was misread, and note business dependencies. Day 10: deliver the report and walk through priorities. Weeks 3 to 8: remediate in stages, starting with the highest-risk items. Week 10 or later: revalidate the fixes and close the findings. Larger or hybrid tenants take longer, and smaller ones can be quicker. The point is to plan the fix and the follow-up as part of the engagement, not as an afterthought. Signs that an audit was done well How do you know you got value? Look for these markers: Findings that name specific users, groups, policies or settings, not general statements. A clear difference between "we saw this," "we could not see this," and "we recommend this." Recommendations that fit your licences and say when an upgrade would be needed. A short executive summary that a non-technical leader can read in five minutes. A remediation list with owners and an order of work. Plain language. If the report needs a glossary to be understood, it will not be acted on. Frequently asked questions Should we run an Office 365 security audit internally or hire someone? Both can work. Internal reviews are cheaper and build knowledge, but they can miss blind spots, and the people who built the tenant sometimes struggle to challenge their own decisions. An independent reviewer brings fresh eyes and credibility with insurers, customers and regulators. Many organisations do a light internal check each quarter and an independent audit each year. What licences do we need for a useful audit? None in particular. An audit can be done on any plan. What changes is the set of recommendations, since features such as Conditional Access, Privileged Identity Management and advanced auditing depend on your plan. A good auditor will separate fixes that use what you already own from changes that need an upgrade. Is an audit the same as a penetration test? No. An audit reviews configuration and evidence against a baseline. A penetration test actively tries to break in. They answer different questions and are often used together. How do we keep audit evidence safe? Store exports and screenshots in a restricted location, remove them when the engagement ends if your agreement says so, and limit who can read the report. Evidence often contains user names and settings that an attacker would find useful. How often should we repeat the audit? At least once a year, plus after any major change such as a merger, tenant migration or licence change. Quick quarterly check-ups on identity and sharing help bridge the gap. Final thoughts A good Office 365 security audit is part technical review and part project management. The technical side needs skill, but the practices that make the biggest difference are the plain ones: a clear question, a written scope, read-only access, evidence you can point to, and findings with names and dates attached. If you are planning your first audit, start small. Choose identity and email, agree the scope, and make sure somebody owns the follow-up. Atlant Security publishes preparation guidance and offers independent reviews for organisations that want a second pair of eyes. Whichever route you choose, the goal is the same: knowing, with evidence, how your tenant really behaves.

Episode details come from the podcast's RSS feed.

Embed this episode

Ready to play

12 Best Practices for an Office 365 Security Audit

0:00 0:13

No transcript yet.

Requests help us decide what to transcribe next.

Frequently asked questions

How long is this episode of PostSphere?

This episode is 0 minutes long.

When was this PostSphere episode published?

This episode was published on October 9, 2026.

Can I download this PostSphere episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.