Guidance
5 min read
Microsoft 365 Security Baseline: Four Controls Beyond MFA
A Microsoft 365 security baseline that goes beyond MFA needs four additional controls: conditional access policies, audit logging retention, data loss prevention rules, and a regular review of privileged admin roles. MFA proves who is signing in. These four controls govern what happens once someone is in, and they are what turns "we use MFA" into a tenant that can actually be audited and defended.
Why MFA Alone Leaves Gaps
MFA raises the cost of a stolen password, but it does not control what a signed-in session can do, how long that session lasts, or what an admin account is allowed to touch. It also does not cover every entry point: legacy authentication protocols, app-level access tokens already issued to connected tools, and service accounts often sit outside normal sign-in flows. Turning off a user's account in the identity provider does not automatically end sessions already open in a browser or revoke tokens already granted to a connected app. Session revocation, app consent review, and deprovisioning in each connected SaaS tool are separate steps, not a side effect of disabling sign-in.
Conditional Access: Control How and Where Sessions Start
Conditional access policies decide whether a sign-in attempt is allowed based on conditions like device compliance, location, or application. A practical starting policy: block legacy authentication protocols tenant-wide, and require a compliant or managed device for anyone signing into an admin portal.
Action: build the policy in report-only mode first, run it for a defined trial period, and review the sign-in logs for accounts that would have been blocked. Verification: use the built-in "what if" policy simulator before enforcing, then confirm a test sign-in from an unmanaged device is actually blocked once you switch to enforce mode. Owner: whoever administers the tenant, with sign-off from IT leadership.
Trade-off: overly strict conditions can lock out legitimate remote staff or break automated service accounts that authenticate without a browser. That is the realistic failure mode, and it is why report-only testing before enforcement matters more than the policy content itself.
Audit Logging: Keep a Record You Can Actually Search
Unified audit logging can be disabled or configured with a retention window shorter than you expect, depending on tenant settings and license tier. A log that is off, or that only covers a short window, is no better than no log when you need to reconstruct what happened.
Action: confirm audit logging is turned on tenant-wide, then decide whether logs need to be exported to a central store so retention is not limited to the default window. Verification: perform a known admin action, such as adding a user to a role, and confirm it appears in a log search afterward. Owner: IT or whoever is accountable for incident readiness; a managed security arrangement can take on the ongoing review so logs are actually watched, not just retained.
Data Loss Prevention: Decide What Should Never Leave
DLP policies watch for defined patterns, like payment card numbers or specific document labels, and act when they appear in outbound email, shared links, or downloads. A workable first policy: flag documents matching a defined sensitive pattern when shared outside the tenant.
Action: deploy new DLP rules in audit or test mode first, so you can see what would have been blocked before you block anything. Verification: send a test file containing a pattern the policy should catch to an external address and confirm it appears in the DLP report. Owner: whoever owns data classification, often shared between IT and a compliance lead; our compliance work covers this kind of policy definition.
Trade-off: enforcing a DLP block before you understand what triggers it can produce false positives that stall normal work and erode trust in the control. Audit mode first, enforce second.
Privileged Role Review: Limit Who Can Change the Tenant
Standing admin access is a liability even with MFA in place, because a compromised admin session can reconfigure the whole tenant. The baseline move is reviewing who holds administrative roles and removing anyone without a current business reason.
Action: list every account holding an admin role, note the business reason for each, and downgrade or remove the ones that no longer have one. Verification: repeat the review on a set schedule and confirm the list has not quietly grown again. Owner: tenant admin, with a second person signing off on removals so no single account makes that call alone.
When to Act, When to Defer, and How to Test It
Act now on conditional access basics and turning on audit logging. Both are largely configuration work rather than new tooling, and the risk of deferring them is open-ended. Defer broad DLP enforcement until you have a working idea of what data actually needs protecting; enforcing rules against undefined "sensitive data" creates noise without reducing risk. Before committing budget or time to any of this, check your current license tier, since some conditional access and DLP capabilities require a higher tier than others, and confirm that in your own tenant rather than assuming.
Test the whole baseline the same way you tested each control: in report-only or audit mode first, with a defined review point before you flip anything to enforce. If a control cannot be tested without disrupting production work, that is a sign it needs a smaller, staged rollout rather than a tenant-wide switch.
A Prioritised Action Checklist
Turn on conditional access in report-only mode for legacy auth blocking and admin device compliance, then review sign-in logs before enforcing (tenant admin).
Confirm unified audit logging is enabled and decide whether logs need export to a central store for longer retention (IT lead, verified by a known-action log search).
Build one DLP policy in audit mode around a defined sensitive data pattern and check the report before enabling blocking (compliance or data owner).
List every admin role holder, confirm a business reason for each, and remove the rest, then repeat the review on a set schedule (tenant admin, second sign-off).
None of these controls replaces a program. They are a baseline, and a baseline needs someone checking it stays in place as the team grows and new apps get connected. If you want a second set of eyes on where your tenant stands today, our virtual CISO work includes this kind of baseline review as part of ongoing security leadership, and a free security review is a low-friction way to see where the gaps actually are before committing further.
Want this looked at for real?
Get a free security review and we will show you where you actually stand.