A compromised Microsoft 365 account rarely stays confined to one inbox. Once an attacker has obtained credentials or a session token, they may create forwarding rules, impersonate finance staff, search SharePoint for sensitive files, register a malicious application or move laterally into connected services. Microsoft 365 hardening is therefore not a single configuration exercise. It is a practical programme for reducing the paths an attacker can use, while keeping the platform workable for employees and support teams.
For UK organisations, the priority is to make sensible, defensible control decisions around identity, email, data and devices. The right design depends on the licences already held, the sensitivity of information, third-party access, workforce patterns and the security tools operating around Microsoft 365.
Microsoft 365 hardening begins with the attack paths
Many tenants have accumulated settings over years of migrations, acquisitions and changing work practices. Guest accounts remain active, legacy authentication is still available for a small number of applications, administrators hold more privilege than they need, and sharing settings differ between Teams, SharePoint and OneDrive. Each exception can be understandable in isolation. Together, they create a difficult-to-govern attack surface.
Start by establishing what is actually connected to the tenant. This should include privileged accounts, external guests, enterprise applications, service accounts, mail flow rules, domains, device management coverage and data-sharing locations. A configuration review should also identify which controls are available under current licensing, rather than assuming a feature is enabled because it appears in Microsoft documentation.
This baseline is valuable for more than technical assurance. It gives security, IT and procurement teams a shared view of the gaps that need native configuration, a specialist security product or a managed service. It also prevents budget being spent on overlapping products while a basic exposure remains unresolved.
Set a risk-based order of work
Not every setting deserves the same urgency. A sensible order is identity first, followed by email security, privileged access, data sharing, endpoint coverage and monitoring. That reflects how many Microsoft 365 compromises begin: a phishing message or stolen password provides a foothold, then weak access controls make that foothold useful.
Regulated organisations may need to apply a different emphasis. A law firm handling confidential client information may place greater weight on external sharing and email encryption. A manufacturer with a distributed workforce may focus first on unmanaged devices and remote access. The principle remains the same: harden the controls that reduce the most credible business risk, not simply those that are easiest to tick off.
Protect identity before adding more tools
Identity is the control plane for Microsoft 365. If an attacker can authenticate as a user or administrator, email filtering and data policies may offer only partial protection.
Make phishing-resistant authentication the standard
Multi-factor authentication should be enforced for all users, with stronger methods prioritised for administrators, finance teams, executives and users handling sensitive data. Authenticator prompts are better than passwords alone, but they are vulnerable to fatigue attacks and social engineering. Where practical, use phishing-resistant methods such as FIDO2 security keys, passkeys or certificate-based authentication.
Conditional Access should then apply context. Policies can require a compliant device for high-risk applications, block sign-ins from unsuitable locations, restrict access from unsupported clients and challenge unusual authentication attempts. Care is required here. Overly broad location blocking can disrupt travellers and legitimate remote staff, while poorly tested device rules can lock out teams during a deployment. Pilot policies with report-only monitoring, assess their impact and retain a documented rollback route.
Reduce privilege and protect the exceptions
Administrators should use separate privileged accounts rather than carrying elevated rights in their daily mailbox identity. Assign the minimum role required, review standing access regularly and use just-in-time elevation where licensing and operational maturity permit it. Global Administrator should be exceptional, not a convenient default.
Emergency access accounts are also necessary, but they are frequently overlooked. Maintain at least two tightly controlled break-glass accounts, protect them with strong credentials and appropriate authentication methods, exclude them only from the policies that could prevent recovery, and monitor every sign-in. Their purpose is resilience, not a permanent bypass.
Disable legacy authentication wherever it is not demonstrably required. Review application registrations and consent permissions as carefully as user accounts. A malicious OAuth application with broad mailbox or file permissions can retain access even after a password reset, so consent governance and app inventory are core hardening controls.
Make email security work with user behaviour
Email remains the most common delivery route for credential theft, payment diversion and malware. Microsoft 365 provides capable native email protection, but the appropriate level of configuration and additional protection varies by organisation.
Defender for Office 365 policies should be tuned for anti-phishing, impersonation, malicious links and attachments, and inbound spam. Configure protection for high-value individuals and domains that are commonly impersonated. Outbound controls matter too: a compromised account sending large volumes of phishing messages should trigger detection and containment quickly.
Domain-based email authentication should be completed for every sending domain. SPF, DKIM and DMARC help reduce spoofing of an organisation's brand, though they do not stop attackers using lookalike domains or convincing social engineering. A DMARC policy should move from monitoring towards enforcement only after legitimate mail sources have been identified and corrected.
Some organisations will benefit from a specialist email security layer, particularly where they face targeted business email compromise, need advanced behavioural detection or want an independent view of Microsoft-native controls. That is not an automatic requirement. Adding another platform introduces administration, alert handling and potential mail-flow complexity. The decision should be based on attack history, internal capability, integration requirements and the measurable detection gap it closes.
Control data without stopping legitimate collaboration
SharePoint, OneDrive and Teams make collaboration easier, but default sharing choices can expose information well beyond the intended audience. Review tenant-wide sharing settings, then examine whether site-level exceptions have created more permissive access. Anonymous links are convenient, but they offer limited assurance over who ultimately receives the file. For sensitive content, authenticated access, expiry dates and restricted download options are usually more proportionate.
Sensitivity labels, encryption and data loss prevention policies can help apply consistent handling to financial information, health data, personal data and commercially confidential documents. Their value depends on policy design and user understanding. A label taxonomy with too many categories will be ignored; a broad rule that blocks routine work will encourage users to find unsafe workarounds.
Begin with a small set of clear classifications and monitor policies before enforcing them. Combine this with regular access reviews for sites, teams and guest users. Security teams should be able to answer a straightforward question: where is sensitive data stored, who can access it, and how is unusual access identified?
Include endpoints, logging and recovery in the design
A secure cloud identity can still be compromised through an unmanaged or infected device. Require device compliance for access to sensitive services where feasible, ensure operating systems and applications are patched, and deploy EDR to provide visibility and response capability. For organisations with a mixed estate, the key question is whether corporate, personally owned and shared devices need different access policies.
Centralise the signals needed to investigate suspicious activity. Authentication logs, audit events, mailbox changes, endpoint telemetry and cloud application alerts should reach a team that can assess and act on them. Retention periods matter. If logs disappear before an incident is discovered, an investigation becomes slower and less certain.
Recovery planning is equally practical. Test the process for disabling accounts, revoking sessions, removing malicious inbox rules, withdrawing application consent and restoring affected data. Confirm that backup arrangements cover the information and recovery objectives that matter to the business. Native retention features and third-party backup are not interchangeable, so assess them against legal, operational and ransomware recovery requirements.
Turn Microsoft 365 hardening into an operating model
Hardening is not complete when the initial policies are deployed. Microsoft changes features, users join and leave, applications request new permissions, and business teams create new collaboration spaces. Assign ownership for policy review, exception approval, access recertification, alert triage and incident response. Keep a centralised risk register that records accepted risks, compensating controls, review dates and accountable owners.
When evaluating additional technology, avoid treating every gap as a reason to buy another platform. Compare tools against the control objective, integration effort, licensing position, reporting needs and the capability of the people who will run them. Independent assessment can be useful where the choice spans email security, identity protection, SaaS security, EDR/XDR or managed detection, because each category can overlap without delivering the same outcome.
The most useful next step is to test one realistic compromise scenario against your tenant: a user approves a phishing prompt, an attacker accesses their mailbox, and then attempts to reach sensitive files. The controls that fail, alert or contain that scenario will show where your Microsoft 365 hardening programme should go next.

