Cloud security architecture review checklist for Azure and Microsoft 365

Latest Comments

No comments to show.
Abstract cloud security architecture review scene with layered identity, access, and governance panels in a restrained purple and gold palette

A cloud security architecture review is not the same as a point-in-time configuration check. For Azure and Microsoft 365, the real question is whether the environment is designed so that identity, data, workloads, and recovery controls still hold up when the business grows, integrates new services, or suffers an incident.

For UK SMEs, that matters because cloud estates often evolve faster than the security model around them. A tenant starts small, then picks up guest access, multiple subscriptions, third-party apps, remote devices, and business-critical collaboration tools. Without periodic review, the architecture can drift away from the original assumptions. If you want a broader view of how architecture reviews fit into the wider design process, our article on secure design reviews and architecture checkpoints in the SDLC is a useful companion.

This checklist is written for security engineers, architects, and analysts who need a practical way to assess Azure and Microsoft 365 together. It focuses on the control planes that matter most: identity, endpoints, data protection, network exposure, application trust, logging, recovery, and governance. It is not a substitute for a full risk assessment, but it is a strong starting point for a defensible review.

Key takeaways

  • Start with identity, because Entra ID and privileged access decisions shape the security of both Azure and Microsoft 365.
  • Treat device posture, data protection, and sharing controls as part of one architecture, not separate admin tasks.
  • Check logging, detection, and recovery together so you can see, contain, and restore from cloud incidents.
  • Rank findings by business impact and implementation effort so remediation work stays practical and achievable.

What this checklist is for and when to use it

Use this checklist when you need to understand whether your cloud security architecture still matches the way the organisation actually operates. Typical triggers include tenant growth, a merger or acquisition, a new managed service provider, a major application rollout, a shift to hybrid working, or an incident that exposed gaps in visibility or access control.

It is also worth using after major platform changes. Examples include enabling new Microsoft 365 workloads, moving workloads into Azure, introducing external collaboration, or changing identity governance. A review at these points helps you catch structural weaknesses before they become operational problems.

A good review should cover more than configuration hygiene. It should answer questions such as: who can access what, from where, using which device, under what conditions, how activity is logged, how quickly you can detect abuse, and how you would recover if a core service or tenant control failed.

Start with scope, ownership, and architecture context

Before looking at controls, define the environment you are reviewing. Map the Microsoft 365 tenant or tenants, Azure subscriptions, management groups, landing zones, and any shared services that sit across business units. Include business-critical services such as email, file storage, collaboration, identity, and customer-facing applications.

It is common to find that the technical estate has outgrown the original ownership model. Make sure you know who owns identity decisions, who owns endpoint policy, who owns cloud platform configuration, and who is responsible for security monitoring and incident response. If those responsibilities are split across IT, infrastructure, and a managed service provider, document the hand-offs clearly.

At this stage, it helps to record the architecture in a simple form. A high-level data-flow diagram, tenant and subscription inventory, and a list of privileged administrative paths are usually enough to expose the main dependencies. If you use a formal architecture method, applying TOGAF for security architecture design can help structure the review without making it overly bureaucratic.

Review identity and access controls first

Identity is the control plane for both Azure and Microsoft 365, so it should be the first area you assess. Start with the Entra ID tenant configuration, including authentication methods, external collaboration settings, and privileged role assignments. Check whether admin accounts are separated from standard user accounts and whether those admin accounts are protected with stronger controls than the rest of the tenant.

Privileged access should be tightly controlled. Review global administrator, privileged role administrator, conditional access administrator, security administrator, and application administrator assignments. Look for standing privileges that could be replaced with just-in-time elevation through Privileged Identity Management. Where possible, require approval, time limits, and strong authentication for privileged activation.

Conditional Access should be treated as a policy layer, not a one-off control. Validate that policies cover all interactive sign-ins, block legacy authentication, and apply stronger requirements to risky sign-ins, unmanaged devices, and privileged roles. If your policies are managed as code, check that the source of truth is version-controlled and that exceptions are tracked. For a deeper implementation view, see implementing Entra ID Conditional Access policies as code.

Multi-factor authentication should be enforced for users and especially for administrators, but the review should go beyond simply asking whether MFA is enabled. Check whether phishing-resistant methods are available for privileged users, whether break-glass accounts are excluded only where strictly necessary, and whether guest access is governed by policy rather than convenience.

Guest and external access is often where architecture starts to weaken. Review cross-tenant access settings, guest lifecycle controls, and whether external users can access the same resources as employees. If guests can reach SharePoint, Teams, or Azure resources, make sure the access model is explicit and time-bound.

Check device and endpoint posture

Identity controls are only as strong as the devices that use them. Review whether corporate devices are enrolled in Intune, whether compliance policies are enforced, and whether Microsoft Defender for Endpoint or an equivalent endpoint protection platform covers the estate. A cloud environment that trusts any device by default is harder to defend and harder to investigate after an incident.

Look at how device compliance is used in Conditional Access. For sensitive applications and data, access should usually depend on a compliant, managed device. If the business allows bring-your-own-device or contractor devices, define what those devices can and cannot do. In many SMEs, the right answer is to allow limited web access but block download, sync, or administrative actions from unmanaged endpoints.

Legacy devices deserve special attention. Older operating systems, unsupported browsers, and shared kiosks can create exceptions that quietly become permanent. If these devices are still needed, isolate them from sensitive workloads and document the compensating controls. If they are not needed, remove them from the trust model.

It is also worth checking whether endpoint posture feeds into your detection strategy. Device compliance failures, risky sign-ins, and endpoint alerts should be visible in the same operational workflow where possible. That makes it easier to correlate identity abuse with endpoint compromise.

Assess Microsoft 365 data protection controls

Microsoft 365 often holds the organisation’s most sensitive business information, so review the data protection model carefully. Start with sensitivity labels. Check whether labels are defined for the main data classes the business actually uses, whether they apply encryption where needed, and whether users understand when to apply them. Labels that exist only on paper do not reduce risk.

Data loss prevention should be aligned to the organisation’s real exposure points. Review whether DLP policies cover email, SharePoint, OneDrive, and Teams, and whether they are tuned to the business’s data types rather than generic patterns. If the organisation handles client records, financial data, or regulated information, the policies should reflect that.

Retention and deletion settings matter as much as protection. Check whether retention policies are consistent with business and operational needs, and whether there are clear rules for preserving evidence after an incident. Over-retention can increase exposure, while under-retention can make investigations and recovery harder.

Sharing controls are another common weak point. Review anonymous link settings, default sharing permissions, external collaboration settings, and whether users can share files outside the organisation without meaningful restrictions. Oversharing is often a design issue rather than a user issue. If you need a practical baseline reference, our article on secure baseline configurations for Microsoft 365 is relevant here.

Mailbox, SharePoint, OneDrive, and Teams should each be checked individually. A secure tenant can still leak data if one workload is configured more permissively than the others. Pay particular attention to guest access in Teams, external sharing in SharePoint, and mailbox forwarding rules that can move data outside the tenant.

Validate Azure network and workload segmentation

In Azure, the main question is whether workloads are exposed more broadly than they need to be. Review the virtual network design, subnet boundaries, network security groups, private endpoints, and any public IP exposure. Public access should be deliberate, documented, and limited to services that genuinely need it.

Segmentation should separate environments and trust levels. Production, test, and development should not share the same flat network unless there is a strong reason. Administrative planes should be isolated from application traffic, and management access should be restricted to hardened paths. If you need a practical pattern for this, designing microsegmentation with Azure network security groups is a useful reference point.

Private endpoints are often a strong improvement over public exposure, but they need to be designed properly. Check DNS resolution, routing, and whether the private path is actually enforced for the services that matter. A private endpoint that is bypassed by a public fallback does not provide the assurance the architecture expects.

Also review how network security is applied between workloads. If one application tier is compromised, can the attacker move laterally to adjacent services, management interfaces, or data stores? Good segmentation reduces blast radius and makes incident containment more realistic.

Review application and workload security design

Applications in Azure and Microsoft 365 should use identity-based trust wherever possible. Review authentication flows, service principals, managed identities, and API permissions. The goal is to remove hard-coded secrets, reduce long-lived credentials, and make service-to-service trust explicit.

Managed identities are usually preferable to stored credentials for Azure resources, but they still need least-privilege role assignments. Check whether the identity has only the permissions required for the workload and whether those permissions are scoped as narrowly as possible. Over-permissioned managed identities are a common blind spot because they are easy to create and harder to notice later.

For applications that integrate with Microsoft 365, review consent settings, app registrations, and delegated permissions. Uncontrolled application consent can create a route for data access that bypasses normal user controls. If your environment has a history of consent sprawl, our article on detecting OAuth consent phishing in Microsoft 365 audit logs shows why this area deserves close attention.

Check secrets handling across the estate. Secrets should be stored in a proper secret store, rotated, and monitored for misuse. Avoid embedding credentials in code, pipelines, or configuration files. Where possible, prefer certificate-based or federated trust over shared passwords or static tokens.

Finally, assess whether the application design follows secure-by-design principles. That means clear trust boundaries, input validation, least privilege, secure defaults, and explicit failure handling. If the architecture relies on every component behaving perfectly, it is probably too brittle for real-world operations.

Confirm logging, detection, and response coverage

A cloud environment without good logging is difficult to operate securely. Review the main telemetry sources across identity, endpoint, Azure control plane, workload activity, and Microsoft 365 audit logs. The question is not just whether logs exist, but whether they are complete enough, retained long enough, and routed to the people who can act on them.

At a minimum, confirm that sign-in logs, audit logs, privileged role activity, endpoint alerts, and Azure resource changes are being collected centrally. If you use Microsoft Sentinel, check the ingestion paths, analytic rules, and whether the most important events are mapped to clear response actions. For a practical detection perspective, see building Microsoft Sentinel analytics rules from MITRE ATT&CK techniques.

Detection coverage should be aligned to realistic attack paths. For example, suspicious admin role activation, impossible travel, consent grants, mailbox forwarding changes, new service principals, and unusual resource creation are all worth prioritising. If you are not sure which events matter most, start with identity abuse, data exfiltration, and privilege escalation paths.

Response ownership matters as much as alert quality. Every high-risk alert should have an owner, an escalation path, and a defined action. If the team cannot tell who triages a critical identity alert at 2am, the architecture is incomplete. Where automation is appropriate, use it for containment steps that are reversible and well understood, not for broad destructive actions.

Test backup, recovery, and resilience assumptions

Cloud services are resilient, but that does not mean your data and configuration are automatically recoverable in the way the business expects. Review backup scope for Microsoft 365 data, Azure workloads, configuration state, and identity dependencies. Make sure the organisation understands what is backed up, what is not, and what the recovery objectives are for each critical service.

Backups should be protected from the same administrative accounts that manage day-to-day operations. Where possible, use immutability, separate credentials, and restricted access to backup systems. Test restores, not just backup success. A backup that cannot be restored within the required time is not a reliable control.

Recovery planning should include tenant-level and platform-level dependencies. If identity services, DNS, key management, or network controls fail, can the business still recover core services? Single points of failure often sit in places that are easy to overlook, so it is worth reviewing them explicitly. Our article on identifying bottlenecks and single points of failure in security architecture is relevant when you are mapping those dependencies.

For Microsoft 365, think through what happens if a privileged account is compromised, a tenant policy is misconfigured, or a malicious change is made to sharing or forwarding rules. For Azure, think through what happens if a subscription is locked down too far, a key vault is unavailable, or a deployment pipeline pushes a bad configuration across multiple environments.

Assess governance, change control, and configuration drift

Cloud security architecture fails quietly when changes are made without revalidation. Review how policies, templates, and exceptions are approved and tracked. If the organisation uses infrastructure as code, policy as code, or configuration baselines, check whether those controls are actually enforced or merely documented.

Good governance means that changes are reviewed against a known baseline, exceptions have an owner and expiry date, and drift is detected quickly. In Azure, that may include Azure Policy, management group structures, and standard landing zone patterns. In Microsoft 365, it may include baseline tenant settings, conditional access policy reviews, and periodic checks on sharing and guest access.

It is also worth checking whether changes are revalidated after deployment. A policy that looked correct in design can behave differently once it meets real users, legacy devices, or third-party integrations. Build a habit of testing the control after the change, not just before it.

Prioritise findings into practical remediation work

Once you have the findings, rank them by business impact, exposure, and effort to fix. The most useful output is not a long list of technical issues, but a short set of actions that reduce risk quickly and a separate set of structural improvements that need roadmap planning.

Quick wins often include tightening guest access, removing stale admin accounts, enforcing stronger Conditional Access, closing unnecessary public exposure, and improving log collection. Structural changes may include redesigning identity governance, reworking network segmentation, introducing private endpoints, or rebuilding the recovery model.

When you present the results, tie each issue to a business outcome. For example, excessive admin privileges increase the chance of tenant-wide compromise, weak sharing controls increase the risk of data leakage, and poor logging increases dwell time and slows response. That framing helps leadership understand why the work matters without turning the review into a compliance exercise.

Common gaps to look for in Azure and Microsoft 365 environments

In practice, the same weaknesses appear again and again. Over-permissioned administrators, inconsistent guest governance, and incomplete logging are among the most common. So are public services that should be private, shared mailboxes or forwarding rules that bypass controls, and recovery plans that have never been tested end to end.

Another frequent issue is inconsistency between workloads. Microsoft 365 may be reasonably well controlled while Azure is left with broader network exposure, or the reverse. The architecture review should look for those mismatches, because attackers usually take the easiest path across the combined environment rather than respecting product boundaries.

If you are building a broader cloud security programme, it can help to align this review with your wider security architecture and resilience work. The aim is not perfection. It is to make sure the environment is understandable, supportable, and resilient enough for the business to operate with confidence.

If you would like help turning this checklist into a practical review plan for your own Azure and Microsoft 365 environment, speak to a consultant.

Frequently asked questions

How often should an Azure and Microsoft 365 security architecture review be carried out?

For most UK SMEs, a formal review makes sense at least annually, with additional reviews after major changes such as tenant growth, mergers, new integrations, large identity changes, or a serious incident. The right cadence depends on how quickly the environment changes and how much business risk sits in the cloud estate.

What is the difference between a security baseline review and a full architecture review?

A baseline review checks whether key settings are configured as expected. A full architecture review goes further and asks whether the overall design still works, including identity flows, trust boundaries, segmentation, logging, recovery, and ownership. In other words, a baseline review looks at settings, while an architecture review looks at the shape and resilience of the whole environment.

Tags:

Comments are closed