Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Microsoft Entra ID, formerly Azure AD, authenticates users to Microsoft 365 and thousands of other applications. Securing it means governing tokens, consent, and roles.

What Is Microsoft Entra ID Security?
Microsoft renamed Azure Active Directory to Microsoft Entra ID in 2023, but the role it plays did not change. For most UK organisations it is now the primary authentication authority—Microsoft 365, hundreds of SaaS applications through single sign-on, Azure resources, and increasingly the Windows devices themselves. It is also internet-facing by design, which makes it a different proposition from the on-premises directory it grew out of.
Entra ID has more than sixty built-in directory roles, and a handful of them are equivalent to owning the tenant. Global Administrator is the obvious one. Less obvious are Privileged Role Administrator, which can grant Global Administrator; Application Administrator and Cloud Application Administrator, which can add credentials to existing service principals and then act as them; and Privileged Authentication Administrator, which can reset the credentials of any user including other administrators.
Microsoft’s own guidance is to keep Global Administrator assignments to fewer than five, and most tenants that have never been reviewed hold considerably more. The structural fix is Privileged Identity Management: make assignments eligible rather than active, require approval and strong authentication at activation, and let them expire. Pair that with cloud-only administrative accounts that have no mailbox and no licence, so that a phished email cannot reach them.
This is the part that is genuinely different from on-premises Active Directory and the part attackers understand best. Any application registered in the tenant has a service principal, and that principal can hold its own credentials—client secrets or certificates—and its own permissions, including application permissions that apply without any user being present.
Two abuses follow. In an illicit consent grant, a user is persuaded to approve an application that requests permissions such as reading all mail or all files. No password is stolen, multi-factor authentication is irrelevant, and the access survives a password reset. Our entry on OAuth consent phishing covers the technique.
In the second, an attacker who has already obtained an administrative role adds a new credential to an existing, trusted, over-permissioned service principal and uses it as quiet persistence. Password resets and session revocations do nothing to it. This was a notable element of the Midnight Blizzard intrusion that Microsoft disclosed in January 2024, where a legacy test tenant account without multi-factor authentication was used to reach an application with elevated access to corporate mailboxes.
Controls: restrict user consent to low-impact permissions with an admin consent request workflow, review all application permissions with broad scopes, alert on credential additions to service principals, and include applications in your access review cycle.
Password attacks against Entra ID are common and largely handled by smart lockout, password protection banned-list policies, and multi-factor authentication. The attacks that succeed now target what comes after authentication: the refresh token. Adversary-in-the-middle phishing kits proxy a genuine sign-in and capture the resulting session, and infostealer malware harvests tokens from browser storage.
Four features address this and all of them need switching on: phishing-resistant authentication for privileged roles, continuous access evaluation so that revocation takes effect in near real time rather than at token expiry, token protection binding sessions to the device, and conditional access requiring a compliant device for sensitive applications. Also block the device code flow except where an inventory shows it is needed, since it has been used repeatedly to obtain tokens without any interactive policy evaluation.
Most organisations synchronise from an on-premises directory, and the synchronisation server is where the two trust boundaries meet. It holds credentials capable of writing to the cloud directory, so it is a Tier 0 asset in exactly the same sense as a domain controller—yet it is frequently patched on a server schedule and administered by whoever manages the virtual machine estate.
Decide the authentication model deliberately: password hash synchronisation, pass-through authentication, or federation. Federation moves the trust to an on-premises service whose signing key, if stolen, allows assertions to be forged for any user, as seen in the Golden SAML technique. Where federation is not required, moving to cloud authentication removes that key from the picture entirely. Where it is required, protect the signing material in hardware and monitor changes to the trust configuration.
Sign-in logs, audit logs, and provisioning logs are retained only briefly in the portal, so export them to a SIEM on day one. The events worth alerting on are: directory role assignments, credential additions to applications and service principals, changes to conditional access policies and their exclusions, new federated domains, consent grants with broad scopes, and any use of a break-glass account. These are rare in normal operation, which makes them cheap detections rather than noise. For UK organisations, this is also the evidence an ICO investigation will ask for after a personal data breach.
At Privia Security, we assess Entra ID tenants and test the paths through them in our cloud security services.
No. Active Directory is an on-premises directory using Kerberos, LDAP, and Group Policy against domain-joined machines. Entra ID is a cloud identity service using OAuth, OpenID Connect, and SAML, with no Group Policy and no organisational unit hierarchy in the same sense. Many organisations run both and synchronise between them, which creates its own security considerations.
Microsoft recommends fewer than five permanent assignments, and the practical target is lower still if you use eligible assignments through Privileged Identity Management. Keep two cloud-only break-glass accounts excluded from conditional access with hardware key protection, and make every other administrative assignment eligible rather than active, with approval and expiry.
It is the identity an application holds in your tenant, with its own credentials and permissions. Because an attacker with sufficient rights can add a new secret or certificate to an existing trusted application and use it as a quiet backdoor, service principals are a favoured persistence mechanism. Alert on credential additions and review application permissions regularly.
Sign-in logs, directory audit logs, and provisioning logs at a minimum, because portal retention is short. Prioritise alerts on role assignments, application credential changes, conditional access policy edits and exclusions, new federated domains, broad consent grants, and break-glass account use. All are rare in normal operation and all appear in real intrusions.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.