Begin with identities and privileged roles
Inventory administrative roles, emergency access accounts, service principals, guest users, and applications with tenant-wide permissions. Assign a business owner to privileged identities and review whether permissions are still required.
Use least privilege and time-bounded privilege activation where available. Separate daily accounts from privileged access, protect administrator sign-ins with strong authentication, and document recovery procedures.
Make conditional access deliberate
Use conditional access to express risk-aware requirements based on user, device, application, and sign-in context. Start in report-only mode where appropriate, test with representative users, and exclude only the narrowly defined emergency accounts required for recovery.
Review legacy authentication exposure and device compliance. Every exception should have a documented reason, an owner, and a review date; otherwise exceptions become invisible permanent access paths.
Govern applications and consent
App registrations and enterprise applications can hold broad API permissions. Limit who may grant tenant-wide consent, review unused applications, and prefer scoped permissions and managed identities over embedded secrets.
Monitor new application consent grants, new credentials on privileged service principals, and changes to federation or authentication methods. A legitimate administrator action can be the first visible step in a compromise.
Collect identity telemetry
Centralize sign-in, audit, and risk-related events with retention appropriate to investigations. Detect unusual privilege changes, repeated authentication failures followed by success, changes to multifactor methods, and suspicious token or session behavior.
Correlate identity signals with endpoint and cloud workload events. No single sign-in anomaly proves compromise; investigation should consider device identity, travel patterns, approved automation, and user context.
Rehearse identity incident response
Document procedures to disable sessions, revoke credentials, rotate secrets, review consent, and preserve audit evidence. A recovery exercise should include what happens when an attacker compromises a privileged account during a cloud-service disruption.
Revisit policies after organizational changes and important platform updates. Control quality depends on continuous ownership, not one successful configuration review.
Practical takeaways
- Review administrator and application privileges.
- Test conditional access and emergency access.
- Monitor consent grants and credential changes.
- Practice token, credential, and session revocation.
Official references and further reading
Publisher pages may change; verify current versions and licensing.