Entra ID Protection: User and Sign-in Risk Policies

What’s the difference between user risk and sign-in risk? Risk detections in Entra ID Protection, risk-based Conditional Access, prerequisites for self-remediation and the review flow.
Technical summary
Technology
Microsoft Entra ID Protection, Conditional Access, SSPR, Password hash sync, Microsoft Graph PowerShell
Topic
Risk-based access policies and risky user review
Level
Intermediate–advanced
Prerequisites
  • A Microsoft Entra ID P2 license
  • The Conditional Access Administrator or Security Administrator role
  • Users’ MFA and SSPR registration
In this article7 sections

The right username, the right password, the expected country. A static Conditional Access policy lets this sign-in through; it can’t tell whether it belongs to an attacker. Microsoft Entra ID Protection closes this gap by assigning a risk level to every sign-in and every user. In this guide we explain the difference between user risk and sign-in risk, the types of risk detection, risk-based Conditional Access design, the prerequisites for self-remediation and the security team’s risky user review flow.

The difference between user risk and sign-in risk

Entra ID Protection works with two separate risk concepts. Mixing them up is the most common cause of wrong policy design in the field.

Feature Sign-in risk User risk
Question Is this sign-in request being made by the account owner? Have this account’s credentials been compromised?
Scope A single sign-in The overall state of the account, cumulative
Typical evidence Anonymous IP, unfamiliar sign-in properties, atypical travel Leaked credentials, a threat intelligence match
Appropriate remediation Proving identity with MFA A secure password change or re-securing the account
When it’s reset For that sign-in, with successful MFA With a password change or an administrator closing the risk

The distinction rests on this: remediating sign-in risk with MFA makes sense, because even if the attacker knows the password, they can’t pass the second factor. User risk is different. If the password has leaked, asking for MFA isn’t enough; the password must change. For an account with high user risk, the possibility that the attacker has registered their own MFA method is also examined separately.

Risk detection types and sources

Detections fall into two groups by timing: real-time detections calculated at sign-in and offline detections produced later by analysis. An offline detection can raise user risk after the sign-in has completed. That’s why the user risk policy kicks in at the next sign-in.

Detection Type Meaning
Leaked credentials User The user’s valid password was found in leak sources
Microsoft Entra threat intelligence User / sign-in Activity matching known attack patterns
Anonymous IP address Sign-in Sign-in from Tor or an anonymizing VPN exit
Unfamiliar sign-in properties Sign-in A location, device or ASN unusual for the user’s history
Atypical travel Sign-in Two locations geographically impossible within a short time
Password spray Sign-in Attempts against many accounts with common passwords
Anomalous token / Token issuer anomaly Sign-in Unusual token properties or traces of a forged token issuer

The full and current list of detections is in Microsoft’s Entra ID Protection documentation; the list changes over time and some detections are shown in detail only at certain license levels. In hybrid environments, password hash sync must be on for leaked credential detection to work. In environments using federation or pass-through authentication, this sync can be enabled in addition. A common situation: in a tenant using PTA, someone asks “why has the leaked credentials detection never fired?” The answer is usually that hash sync is off. For choosing a sync method, see the Entra Connect Sync vs Cloud Sync comparison.

Risk-based Conditional Access policies

Risk policies are built in Conditional Access with the User risk and Sign-in risk conditions. Microsoft recommends using Conditional Access instead of the legacy user risk and sign-in risk policies in the Identity Protection screen. If there are legacy policies in the environment, move them; don’t manage both at the same time, because two overlapping policies turn troubleshooting into a nightmare. For the general design framework, see the Conditional Access planning guide.

Two policies are enough to start:

Policy Condition Access control Sign-in
Sign-in risk Sign-in risk: High, Medium Require MFA or authentication strength Sign-in frequency: Every time
User risk User risk: High Require MFA + Require password change Sign-in frequency: Every time

Design notes:

  • Scope: All users and all cloud apps. Emergency (break-glass) accounts are excluded; but every sign-in with these accounts should raise a separate alarm.
  • Choosing the user risk level: Starting with High only lets you see the false positive rate and the help desk load. Medium is added after measuring.
  • Passwordless users: For a user with no password, or who signs in only with FIDO2/Windows Hello for Business, a “change password” remediation is meaningless. Check the risk remediation options Microsoft offers for this scenario in the current documentation.
  • Phishing-resistant method: In the sign-in risk policy, requiring a phishing-resistant authentication strength instead of just “Require MFA” stops the attacker from passing MFA through a proxy server in an AiTM attack. Push approval doesn’t stop AiTM. For details, see the FIDO2 and passkey migration plan.
  • Start with report-only: Run the policies in report-only mode for a week or two, review the “Report-only” results in the sign-in logs, then turn them on.

Self-remediation and license requirements

Risk-based Conditional Access policies and full detailed reports on risky users require a Microsoft Entra ID P2 license (included in bundles such as Microsoft 365 E5). In P1 and the free tier, reporting is limited and the risk condition can’t be used in policies. Verify the current license mapping on Microsoft’s licensing page.

Self-remediation works if the user has been prepared in advance. Without preparation, the policy leaves the user locked out:

  1. MFA registration: A user without MFA registration at the moment of risk is blocked; allowing them to register at that moment could let the attacker register their own method. The registration campaign must finish before the risk policies.
  2. Password writeback: According to Microsoft’s documentation, the secure password change in risk remediation (MFA followed by a password change) doesn’t require self-service password reset (SSPR) to be on. For a hybrid user’s password changed in the cloud to be written to on-premises AD, however, password writeback must be enabled in Entra Connect. If writeback is off, the user ends up with a new password in the cloud and the old one on-premises.
  3. On-premises password change: In environments using password hash sync, the “allow on-premises password change to reset user risk” setting lets a change the user makes in AD close the risk.

The risky user review flow

Automatic remediation doesn’t close every case. An analyst should see high-risk events, including risks the user remediated themselves. The order of checks:

  1. Triage: Open the user in the Risky users report; review the risk history, the detections that triggered it and the related sign-ins.
  2. Context: IP address, device ID, application, successful/failed MFA, sign-ins to other accounts from the same IP.
  3. Persistence check: MFA methods added in recent days, new mailbox rules and forwarding, app consents (OAuth consent), newly registered devices. The attacker leaves the door ajar first.
  4. Decision: If it’s a real compromise, Confirm user compromised; revoke the user’s sessions (Revoke sessions), reset the password, and remove suspicious MFA methods and rules. If it’s a false positive, Confirm safe or Dismiss user risk.
  5. Record: The reasoning for the decision is written into the incident record.

Microsoft Graph PowerShell can be used for bulk review and automation:

Connect-MgGraph -Scopes "IdentityRiskyUser.ReadWrite.All"

# High-risk users still in the at-risk state
Get-MgRiskyUser -Filter "riskLevel eq 'high' and riskState eq 'atRisk'" |
  Select-Object UserPrincipalName, RiskLevel, RiskState, RiskLastUpdatedDateTime

# After review: confirm compromise or dismiss the risk
Confirm-MgRiskyUserCompromised -UserIds "<user-object-id>"
Invoke-MgDismissRiskyUser -UserIds "<user-object-id>"

Exporting risk detections and risky user events to Microsoft Sentinel or the SIEM you use lets you correlate identity events with endpoint and email events. Sign-in and audit logs are retained in the portal for a limited time. When an incident is noticed three weeks later, the log may already be gone; plan the export to Log Analytics or the SIEM from the start for long-term investigation.

False positives and tuning

The risk engine learns the user’s past behavior; in new tenants and for new users, seeing more “unfamiliar properties” detections in the first weeks is expected. Ways to reduce the noise:

  • Named locations: Mark the company’s egress IPs as trusted; this information feeds into the risk calculation.
  • Corporate VPN and proxy: In architectures that send all traffic out through a single cloud proxy, location-based detections can be misleading; define the proxy egress IPs.
  • Feedback: The analyst’s Confirm safe and Confirm compromised decisions feed the model. Risks left unreviewed both create noise and block learning. Don’t let the queue pile up.
  • Service accounts: For automation accounts that don’t sign in interactively, workload identities and managed identities should be preferred over user risk.

Evaluating risk policies together with the existing Conditional Access and MFA design falls within Microsoft 365 security work, and tying risk-based access decisions into the organization’s overall identity strategy falls within identity and access management work.

Frequently asked questions

Which license does Entra ID Protection require?

Using risk conditions in Conditional Access policies and seeing detailed reports on risky users requires Microsoft Entra ID P2. The free and P1 tiers have only limited reporting. Verify the current mapping in Microsoft’s licensing documentation.

Which risk levels should I select in the sign-in risk policy?

A common starting point is requiring MFA at the Medium and High levels. Adding Low can noticeably increase false positives and user friction; measure its impact in report-only mode first.

If the user remediated their own risk, is a review still needed?

For high-risk events, yes. Self-remediation changes the password but doesn’t clean up MFA methods, mailbox rules or app consents added by the attacker. An analyst checks for this persistence.

Does leaked credential detection work in a hybrid environment?

For it to work, password hash synchronization must be enabled with Entra Connect or Cloud Sync. If federation or pass-through authentication is used and password hash sync is off, leaked credentials can’t be compared; sync can be turned on alongside these methods too.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Cybersecurity

Firewall, VPN, endpoint security and threat prevention. We reduce the attack surface and measure and report the risks.

Knowledge Center

Other articles