An account without MFA, a legacy protocol left enabled or a mailbox rule forwarding mail outside can be enough for a Microsoft 365 account to be compromised. Which of the security features included in the license are used varies from environment to environment; turning on every feature isn’t the right decision for every environment either. The list below collects the settings to review in order of priority. Try every change in a pilot group first and inform users in advance; an MFA requirement switched on without warning brings a flood of calls to the help desk.
Priority 1: Identity
In Microsoft 365 the security boundary isn’t the network but identity. When a password leaks, the only thing standing in the attacker’s way is authentication policies. That’s why the first job starts here.
- MFA for all users. If there’s no Entra ID P1 license, the free security defaults should be enabled; if there’s P1, it should be managed with Conditional Access. The two aren’t used at the same time.
- Blocking legacy authentication. Legacy protocols that don’t support MFA are the main route for password spraying attacks.
- Administrator roles. Reduce the number of Global Administrators (Microsoft recommends no fewer than two and no more than five), separate administrator accounts from everyday accounts and use strong MFA.
- Emergency accounts. Microsoft recommends keeping at least two emergency (break-glass) accounts and monitoring these accounts separately.
- Application consent. Limit users’ ability to grant third-party applications consent to organizational data.
Common identity mistakes
MFA appears to be “on” but is assigned to only some users; the legacy per-user MFA setting and Conditional Access policies are used together and nobody knows which one applies. Another: the administrator account is also the everyday account used to read email and browse the internet. When this account gets phished, the whole tenant falls.
Emergency accounts may be missing, or their password may be sitting in an employee’s personal notes. According to Microsoft’s guidance, these accounts are excluded from Conditional Access policies that block or restrict sign-in, protected with a phishing-resistant method such as passkeys (FIDO2) or certificate-based authentication, and their credentials are stored in a physically secure place. A separate alert should be defined for each of their sign-ins. If someone signs in to an emergency account that isn’t used, that’s an incident.
Application consent is a quiet risk. When a user clicks accept on the “This app wants to read your mailbox” screen, persistent access is granted without a password. Limit user consent to verified publishers and low-risk permissions, and enable the admin consent workflow for the rest.
Priority 2: Email
Email is both the initial entry point and the channel most abused once an account is compromised. Most settings on the Exchange Online side take a few minutes; their effect is long-term.
- Block automatic forwarding to external recipients; it’s one of the most common behaviors seen in compromised accounts.
- Configure SPF, DKIM and DMARC, and tighten DMARC in stages.
- Enable impersonation protection in anti-phishing policies for executives and the finance team.
- If you have a Defender for Office 365 license, enable Safe Links and Safe Attachments policies.
The typical DMARC mistake is publishing the record and forgetting it at p=none for years. This level only produces reports; it doesn’t stop spoofed mail. The right way: first read the reports and align all legitimate services sending mail on behalf of the organization (invoicing system, CRM, newsletter tool) with SPF and DKIM, then move to quarantine and then reject. A DMARC record whose reports nobody reads is just decoration.
On the forwarding side, disabling external forwarding alone isn’t enough. In incident investigations, look at inbox rules: rules that delete emails containing words like “invoice”, “payment” or “bank”, or move them to an archive folder, are the classic trace of a compromised account.
Priority 3: Devices and sharing
Even if identity is protected, a session token taken from an insecure device or an anonymous sharing link carries data out. The second line of defense is device and sharing policies.
- Define device enrollment and compliance policies (encryption, up-to-date operating system, screen lock) with Intune; tie access to corporate data to a compliant device condition.
- On personal devices, keep corporate data inside the app with app protection (MAM) policies.
- Limit the SharePoint and OneDrive external sharing level to the organization’s needs; put an expiry on anonymous links.
Writing a compliance policy in Intune is half the job. If you don’t plan when a non-compliant device counts as non-compliant (grace period), how the user is notified and how this ties into Conditional Access, the policy either blocks nothing or one morning leaves the sales team unable to reach their mail from their phones. Reporting first, then a pilot group, then general rollout.
In sharing, set the default link type to “people in your organization” or “specific people” instead of “anyone”. If there’s a business need for anonymous links, set an expiry. An anonymous link with no expiry is a key whose holder you don’t know.
Priority 4: Visibility and data
Without logs, an incident investigation turns into guesswork. Verify that logging is on today, not during an incident.
- Verify that the unified audit log is enabled.
- Define alerts for administrator role changes, MFA method changes and suspicious mailbox rules.
- Run starter policies for sensitivity labels and DLP in audit mode.
- Taking Microsoft’s shared responsibility model into account, evaluate your backup strategy for deleted or encrypted data.
Check who receives the alerts. A common situation: alerts go to the address of an employee who left years ago, or to a shared mailbox nobody reads. Defining an alert doesn’t mean it gets read.
The retention period of audit logs varies by license. Because incidents are often noticed weeks later, check whether the retention period meets your needs; if it doesn’t, export the logs to a SIEM. With DLP and labels, don’t go straight to blocking mode. Run them in audit mode for a few weeks to see which processes carry sensitive data; otherwise, on the first day, accounting’s bank file submission stops.
Backup is often misunderstood. Microsoft is responsible for the availability of the service; data that’s accidentally deleted, encrypted by ransomware or past its retention period is your responsibility. The recycle bin and version history aren’t backups.
A note on licensing
Some of the controls above exist in every tenant; some require a license. Before you start planning, clarify what your license covers.
| Feature | Required license (summary) |
|---|---|
| Security defaults | All tenants, free |
| Conditional Access | Microsoft Entra ID P1 (e.g. Microsoft 365 Business Premium, E3) |
| Risk-based policies, PIM | Microsoft Entra ID P2 (e.g. E5) |
License contents are changed by Microsoft from time to time; verify the current situation in Microsoft Learn and the licensing documentation.
Short checklist for the implementation order
- Emergency accounts have been created, tested and excluded from policies.
- MFA is mandatory for all users through a single method (security defaults or Conditional Access).
- Legacy authentication has been checked in the sign-in logs and disabled.
- External forwarding is off, and DMARC reports are being read.
- The audit log is on, and alerts go to an active mailbox.
This list isn’t a one-off job to complete and close. Settings drift as new applications, new licenses and new employees arrive. Review it every three months.
For a comprehensive assessment of your tenant, see our Microsoft 365 Security page; for identity processes, our Identity and Access Management page; and for the email side, our Email Security page. Once identity and sharing boundaries are set, design user groups and portal access accordingly during SharePoint intranet setup.