A successful sign-in does not always mean a legitimate sign-in.

An attacker using a valid password, a stolen session token, an anonymous proxy, or credentials exposed in a breach can look surprisingly normal if you only review whether authentication succeeded or failed.

That is why risky sign-ins in Microsoft Entra ID deserve their own monitoring and investigation workflow.

Microsoft Entra ID Protection evaluates sign-in activity and assigns risk signals that can help identify authentication attempts that may not have been performed by the legitimate user. Those signals can include unfamiliar sign-in properties, anomalous travel patterns, suspicious or malicious IP addresses, leaked credentials, token-related anomalies, and other indicators Microsoft associates with identity compromise.

The challenge is not simply finding a red icon in the Entra portal. The real work is determining:

  • Which risky sign-ins actually matter?
  • Which users should be investigated first?
  • What does low, medium, or high sign-in risk mean operationally?
  • How can you detect risky sign-ins with KQL instead of checking the portal manually?
  • How should Conditional Access respond when a risky authentication occurs?

This guide walks through a practical approach to detecting, investigating, and responding to risky sign-ins in Microsoft Entra ID.

Quick Answer: How Do You Detect Risky Sign-ins in Entra ID?

The fastest manual method is to use the Risky sign-ins report in Microsoft Entra ID Protection.

For continuous monitoring, send Microsoft Entra sign-in logs to a Log Analytics workspace or Microsoft Sentinel and query the SigninLogs table.

A simple starting query is:

SigninLogs
| where TimeGenerated > ago(30d)
| where RiskLevelDuringSignIn in ("medium", "high")
| project TimeGenerated,
          UserPrincipalName,
          AppDisplayName,
          IPAddress,
          Location,
          RiskLevelDuringSignIn,
          RiskLevelAggregated,
          RiskState,
          RiskEventTypes_V2,
          ConditionalAccessStatus
| order by TimeGenerated desc

This gives you a practical list of sign-ins that Microsoft Entra evaluated as medium or high risk during authentication.

For a broader operational view, also review:

  • RiskState == "atRisk"
  • IsRisky == true
  • AADUserRiskEvents if you export ID Protection risk detections
  • MFA results and Conditional Access outcomes
  • Device, application, location, IP address, and user-agent context

What Is a Risky Sign-in in Microsoft Entra ID?

A risky sign-in is an authentication event that Microsoft Entra ID Protection believes may not have been performed by the legitimate identity owner.

During sign-in, ID Protection analyzes a large set of signals and calculates a sign-in risk level that represents the probability that the authentication request is unauthorized.

This is different from user risk.

| Signal | What it represents | Example question | | --- | --- | --- | | Sign-in risk | The probability that a specific authentication request is not legitimate | Was this sign-in performed by the real user? | | User risk | The probability that the user account itself is compromised | Is this identity currently compromised? | | Risk detection | A specific signal or event that contributed to the risk decision | What suspicious behavior did Microsoft detect? |

A single risky sign-in can contribute to user risk, but the two concepts should not be treated as interchangeable.

For security operations, that distinction matters because your response to a suspicious sign-in can be different from your response to an account that Microsoft believes is actively compromised.

Common Microsoft Entra Risk Detections

Microsoft Entra ID Protection can identify multiple types of suspicious activity. The exact detections available depend on licensing and Microsoft’s current detection capabilities, but commonly observed categories include:

  • Unfamiliar sign-in properties
  • Anonymous IP address activity
  • Malicious or suspicious IP address activity
  • Atypical or unlikely travel patterns
  • Leaked credentials
  • Threat-intelligence detections
  • Token-related anomalies
  • Activity associated with malware or known malicious infrastructure

Some detections are evaluated in real time, while others are calculated offline after Microsoft has more context.

That means a sign-in that initially appears low risk or normal can later be reclassified when additional telemetry is processed.

For that reason, a mature Entra ID security monitoring process should review both real-time sign-in risk and aggregated risk.

Understanding the Risk Fields in SigninLogs

If you send Microsoft Entra sign-in logs to Log Analytics, the SigninLogs table contains several fields that are especially useful for identity-security monitoring.

| Field | Why it matters | | --- | --- | | IsRisky | Indicates whether the sign-in is considered risky | | RiskLevelDuringSignIn | Risk level calculated during authentication | | RiskLevelAggregated | Aggregated risk after additional analysis | | RiskState | Current state such as atRisk, remediated, dismissed, or confirmedCompromised | | RiskDetail | Additional information about remediation or risk state | | RiskEventTypes_V2 | Risk detections associated with the sign-in | | IPAddress | Source IP used for the authentication | | Location / LocationDetails | Geographic context for the sign-in | | AppDisplayName | Application the user attempted to access | | ConditionalAccessStatus | Whether Conditional Access succeeded, failed, or was not applied | | AuthenticationDetails | Authentication-step details that can help validate MFA behavior | | DeviceDetail | Device context associated with the sign-in |

The values for RiskLevelDuringSignIn and RiskLevelAggregated can include none, low, medium, high, or hidden.

If risk details appear as hidden, that can indicate the tenant does not have the Microsoft Entra ID Protection licensing required to expose full risk information.

Method 1: Detect Risky Sign-ins in the Microsoft Entra Admin Center

For an immediate manual review, start with Microsoft Entra ID Protection.

  1. Sign in to the Microsoft Entra admin center.
  2. Browse to Entra ID > Protection > Risky activities > Risky sign-ins.
  3. Filter by Risk level, Risk state, user, application, location, or date.
  4. Open suspicious events and review the sign-in details.
  5. Correlate the event with the user’s normal behavior, device history, Conditional Access result, and MFA activity.

The Risky sign-ins report can show useful investigation context including:

  • User
  • Application
  • Sign-in date and time
  • IP address
  • Location
  • Device information
  • MFA details
  • Conditional Access policies applied
  • Risk level
  • Risk state
  • Risk detections associated with the event

A good starting filter is medium and high risk sign-ins that are still at risk.

Those are typically more actionable than a large list that includes already-remediated events.

Method 2: Detect Risky Sign-ins with KQL in Log Analytics

For repeatable monitoring, KQL is much more useful than manually checking the portal.

Find Medium- and High-Risk Sign-ins

SigninLogs
| where TimeGenerated > ago(30d)
| where RiskLevelDuringSignIn in ("medium", "high")
| project TimeGenerated,
          UserPrincipalName,
          AppDisplayName,
          IPAddress,
          Location,
          RiskLevelDuringSignIn,
          RiskLevelAggregated,
          RiskState,
          RiskEventTypes_V2,
          ResultType,
          ConditionalAccessStatus
| order by TimeGenerated desc

This is one of the most useful baseline queries for an Entra risky sign-in dashboard.

Find Sign-ins That Are Still At Risk

SigninLogs
| where TimeGenerated > ago(30d)
| where RiskState == "atRisk"
| project TimeGenerated,
          UserPrincipalName,
          AppDisplayName,
          IPAddress,
          Location,
          RiskLevelDuringSignIn,
          RiskLevelAggregated,
          RiskEventTypes_V2,
          ConditionalAccessStatus
| order by TimeGenerated desc

This helps remove some noise from events that have already been remediated or dismissed.

Find High-Risk Successful Sign-ins

A failed risky sign-in matters, but a successful high-risk sign-in deserves immediate attention because access may already have been granted.

SigninLogs
| where TimeGenerated > ago(30d)
| where RiskLevelDuringSignIn == "high"
| where ResultType == "0"
| project TimeGenerated,
          UserPrincipalName,
          AppDisplayName,
          IPAddress,
          Location,
          RiskState,
          RiskEventTypes_V2,
          AuthenticationRequirement,
          ConditionalAccessStatus
| order by TimeGenerated desc

If you only have time to review one risky-sign-in queue, this is a strong candidate.

Summarize Risky Sign-ins by User

SigninLogs
| where TimeGenerated > ago(30d)
| where RiskLevelDuringSignIn in ("medium", "high")
| summarize RiskySignIns = count(),
            HighRisk = countif(RiskLevelDuringSignIn == "high"),
            IPAddresses = make_set(IPAddress, 20),
            Locations = make_set(Location, 20),
            Apps = make_set(AppDisplayName, 20)
  by UserPrincipalName
| order by HighRisk desc, RiskySignIns desc

This is useful for identifying users generating repeated risky authentication activity.

Find Risky Sign-ins from Multiple Countries

SigninLogs
| where TimeGenerated > ago(7d)
| where RiskLevelDuringSignIn in ("medium", "high")
| extend Country = tostring(LocationDetails.countryOrRegion)
| summarize SignInCount = count(),
            Countries = make_set(Country, 20),
            IPAddresses = make_set(IPAddress, 20)
  by UserPrincipalName
| where array_length(Countries) > 1
| order by SignInCount desc

Multiple countries do not automatically mean compromise. VPNs, mobile carriers, corporate egress points, travel, and cloud security services can all affect geolocation.

But as an investigation pivot, country changes are useful when combined with other risk signals.

Method 3: Query Microsoft Entra ID Protection Risk Events

If you export ID Protection data through diagnostic settings, you can query the AADUserRiskEvents table in Log Analytics.

This table is valuable because it focuses on the specific risk detections associated with users and sign-ins.

Find Active Risk Detections

AADUserRiskEvents
| where TimeGenerated > ago(30d)
| where RiskState == "atRisk"
| project TimeGenerated,
          UserPrincipalName,
          RiskEventType,
          RiskLevel,
          RiskState,
          RiskDetail,
          IPAddress,
          Location
| order by TimeGenerated desc

Find High-Risk Detection Types

AADUserRiskEvents
| where TimeGenerated > ago(30d)
| where RiskLevel == "high"
| summarize Detections = count() by RiskEventType
| order by Detections desc

This can help you understand why your tenant is generating high-risk identity events instead of treating every risky sign-in as the same problem.

Which Risky Sign-ins Should You Investigate First?

Not every risky sign-in deserves the same priority.

I would generally investigate in this order.

1. High-Risk Successful Sign-ins

A successful authentication with high sign-in risk is one of the clearest indicators that an attacker may already have gained access.

Validate:

  • Was MFA performed?
  • Was MFA satisfied by a previous claim or existing session?
  • Was Conditional Access applied?
  • Is the device known and compliant?
  • Is the IP address expected?
  • Does the user recognize the activity?
  • Were additional applications accessed after the sign-in?

2. Risky Sign-ins for Privileged Accounts

A medium-risk event for a Global Administrator, Privileged Role Administrator, Security Administrator, or Azure subscription owner can be more important than a high-risk event for a low-impact account.

Risk level should always be interpreted together with identity privilege and business impact.

3. Repeated Risky Sign-ins from the Same User

A single anomalous event may be explainable.

A sequence of risky sign-ins across multiple IPs, countries, applications, or devices is much harder to dismiss.

Look for clusters instead of isolated alerts.

4. Risky Sign-ins Where Conditional Access Did Not Apply

If a risky sign-in succeeded and ConditionalAccessStatus indicates that policies were not applied, investigate why.

Possible causes include:

  • Policy exclusions
  • Application scope gaps
  • User or group exclusions
  • Unsupported authentication flows
  • Emergency-access accounts
  • Legacy or unusual client behavior

A risky sign-in can expose a detection problem, but it can also expose a policy coverage problem.

5. Sign-ins Associated with High-Confidence Risk Detections

Some risk detections deserve immediate escalation, especially when they indicate exposed credentials, malicious infrastructure, token abuse, or threat-intelligence matches.

Do not rely only on the words “medium” or “high.” Review the detection type.

A Practical Risky Sign-in Investigation Workflow

When you identify a risky sign-in, avoid jumping directly to “block the user” or “dismiss the alert.”

Use a repeatable investigation process.

Step 1: Validate the Identity and Time

Confirm:

  • User principal name
  • Timestamp
  • Application accessed
  • Authentication result
  • Risk level and state

Step 2: Review the Source

Check:

  • IP address
  • Country and region
  • ISP or hosting-provider context if available
  • VPN or proxy usage
  • Whether the IP appears in other users’ sign-ins

A single suspicious IP used by many accounts can indicate password spraying, shared infrastructure, or a corporate egress point.

Step 3: Review Device and Client Context

Compare the event to the user’s normal pattern.

Ask:

  • Is the device registered?
  • Is it compliant?
  • Is the operating system expected?
  • Is the browser or client application normal for the user?
  • Is the user agent unusual?

Step 4: Validate MFA and Conditional Access

This is critical.

A sign-in can be risky but still successfully self-remediate if a risk-based Conditional Access policy requires MFA and the legitimate user completes it.

Review:

  • Authentication requirement
  • Authentication details
  • MFA result
  • Conditional Access policies applied
  • Grant controls
  • Session controls

Step 5: Look for Activity After the Sign-in

If you suspect compromise, determine what happened next.

Review:

  • Additional sign-ins
  • Azure portal activity
  • Microsoft 365 activity
  • Role changes
  • OAuth consent events
  • Application registrations
  • Mailbox activity
  • Security-setting changes
  • New authentication methods

The risky sign-in may be the entry point, not the entire incident.

Step 6: Confirm with the User Carefully

If appropriate, contact the user through a trusted channel and ask whether the sign-in is expected.

Do not rely blindly on email or Teams if the account itself may be compromised.

Provide enough context for the user to recognize the event, such as time, application, device, and approximate location.

How Conditional Access Can Automatically Respond to Sign-in Risk

Detection is useful. Automated protection is better.

Microsoft Entra Conditional Access can use sign-in risk as a policy condition.

Common responses include:

  • Require multifactor authentication
  • Require reauthentication
  • Block access
  • Allow the user to self-remediate risk through a protected authentication flow

A common design is to require MFA for medium- and high-risk sign-ins, while reserving stronger restrictions or blocking behavior for the highest-risk scenarios based on organizational tolerance.

The exact policy should be tested carefully because risk-based Conditional Access can affect legitimate users during travel, VPN changes, device migrations, and unusual access patterns.

Use report-only mode and staged deployments before broad enforcement where appropriate.

Common Mistakes When Monitoring Risky Sign-ins

Mistake 1: Reviewing Only Failed Sign-ins

A failed attack is useful telemetry.

A successful risky sign-in is usually more urgent.

Do not filter your investigation queue to failures only.

Mistake 2: Treating Location as Proof of Compromise

Geolocation is evidence, not proof.

VPNs, mobile networks, cloud proxies, secure web gateways, and corporate egress services can make a legitimate user appear to be somewhere unexpected.

Use location together with device, application, authentication, IP, and behavior context.

Mistake 3: Ignoring Remediation State

A risky sign-in that was remediated through MFA should not be treated the same way as an unresolved high-risk sign-in that successfully accessed a sensitive application.

Filter on RiskState so your operational queue reflects what still needs attention.

Mistake 4: Looking at Risk Without Looking at Privilege

A security team that sorts only by risk level can miss the bigger business impact.

A medium-risk sign-in to a privileged administrator may deserve faster escalation than a high-risk event for an account with minimal access.

Mistake 5: Manually Checking the Portal Once a Month

Risky sign-ins are time-sensitive.

The faster you detect compromise, the less time an attacker has to establish persistence, register applications, modify authentication methods, or move laterally into cloud resources.

Use Log Analytics, Microsoft Sentinel, workbooks, scheduled analytics rules, or other automation to make identity-risk monitoring continuous.

Building a Risky Sign-in Dashboard

A useful Microsoft Entra risky sign-in dashboard should show more than a count of alerts.

I would include at least:

  • Total risky sign-ins over the selected period
  • High-risk successful sign-ins
  • Medium- and high-risk sign-ins by user
  • Risky sign-ins by country
  • Top source IP addresses
  • Top applications involved
  • Risk detections by type
  • Risky sign-ins for privileged users
  • Unresolved atRisk events
  • Conditional Access results
  • MFA outcomes
  • Trend of risky sign-ins over time

This turns raw Entra ID logs into an operational view that can answer the question:

Where should I investigate first?

How AzureWise Helps Detect Risky Sign-ins

Risky sign-in visibility is one of the core reasons I built the AzureWise Identity Security Assessment.

The AzureWise workbook and KQL query pack are designed to help Microsoft Entra and Azure administrators quickly review identity-security signals without rebuilding the same queries and visualizations from scratch.

The assessment includes visibility into areas such as:

  • Risky sign-ins
  • MFA coverage
  • Legacy authentication
  • Impossible travel and suspicious locations
  • Privileged role changes
  • Recent privileged activity
  • Service principal sign-ins
  • OAuth consent activity
  • Guest-user activity

AzureWise is designed as a read-only security assessment toolkit. It uses data already available in your Azure and Microsoft Entra environment to help surface potential identity-security gaps and investigation priorities.

Want a faster way to review Microsoft Entra ID risk?

The AzureWise Identity Security Assessment is available as a one-time $99 license and includes the workbook, KQL query pack, and setup documentation.

>

View the AzureWise Identity Security Assessment | See How It Works

Risky Sign-in Detection Checklist

Use this checklist when reviewing your Microsoft Entra environment:

  • [ ] Review medium- and high-risk sign-ins.
  • [ ] Prioritize successful high-risk authentications.
  • [ ] Identify unresolved events where RiskState == "atRisk".
  • [ ] Review risky sign-ins for privileged identities.
  • [ ] Correlate risk with MFA and Conditional Access results.
  • [ ] Review IP address, location, device, application, and user-agent context.
  • [ ] Look for repeated risky events involving the same user.
  • [ ] Review AADUserRiskEvents if ID Protection risk events are exported.
  • [ ] Confirm that risk-based Conditional Access policies cover the users and applications you expect.
  • [ ] Send Entra sign-in logs to Log Analytics or Microsoft Sentinel for continuous monitoring.
  • [ ] Create alerts for high-risk successful sign-ins.
  • [ ] Track risky sign-in trends over time instead of relying on one-time reviews.

FAQ: Risky Sign-ins in Microsoft Entra ID

What is a risky sign-in in Microsoft Entra ID?

A risky sign-in is an authentication attempt that Microsoft Entra ID Protection has evaluated as potentially unauthorized. The service uses signals such as sign-in behavior, IP reputation, credential exposure, location patterns, threat intelligence, and other telemetry to calculate risk.

What is the difference between sign-in risk and user risk?

Sign-in risk applies to a specific authentication event. User risk represents the probability that the user account itself is compromised. A risky sign-in can contribute to user risk, but the two values describe different security conditions.

How do I find risky sign-ins with KQL?

Query the SigninLogs table and filter on fields such as RiskLevelDuringSignIn, RiskLevelAggregated, RiskState, or IsRisky.

For example:

SigninLogs
| where RiskLevelDuringSignIn in ("medium", "high")

What does RiskState atRisk mean?

atRisk indicates that the risky user, sign-in, or risk event still has active risk associated with it and has not been fully remediated, dismissed, or confirmed safe.

What is the difference between RiskLevelDuringSignIn and RiskLevelAggregated?

RiskLevelDuringSignIn reflects the risk calculated during the authentication event. RiskLevelAggregated can reflect additional analysis that occurs after the sign-in as Microsoft processes more signals. The two values may therefore differ.

Why does Microsoft Entra show risk as hidden?

Some detailed Microsoft Entra ID Protection risk information requires Microsoft Entra ID P2 licensing. In tenants without the required licensing, risk fields can be returned as hidden instead of exposing the underlying details.

Can Conditional Access automatically respond to risky sign-ins?

Yes. Conditional Access can use sign-in risk as a policy condition and require controls such as MFA, reauthentication, risk remediation, or blocking access depending on your configuration.

How long are risky sign-ins visible in the Entra report?

Microsoft documents that the Risky sign-ins report provides filterable data for up to approximately the previous 30 days. For longer retention and custom analysis, export the relevant logs to Log Analytics, Microsoft Sentinel, or another SIEM platform.

Should every risky sign-in be treated as a security incident?

No. Risk detections are investigation signals, not automatic proof of compromise. Prioritize high-risk successful sign-ins, privileged identities, repeated suspicious patterns, unresolved risk, and events where Conditional Access or MFA controls did not behave as expected.

Final Takeaway

The most important lesson is simple:

Do not treat authentication success as proof that the user is legitimate.

Microsoft Entra risky sign-ins give you a way to detect authentication events that look wrong even when the credentials are correct.

Start with the Risky sign-ins report for manual investigation, then move the monitoring into Log Analytics or Microsoft Sentinel so you can continuously detect medium- and high-risk sign-ins with KQL.

Prioritize successful high-risk authentication, privileged users, unresolved risk, repeated suspicious behavior, and events where MFA or Conditional Access did not provide the protection you expected.

The goal is not to eliminate every risk alert.

The goal is to identify the small number of risky sign-ins that could represent real account compromise before they turn into a larger cloud security incident.

Learn more about the AzureWise Identity Security Assessment

Microsoft References

Put the guidance into practice

Want the Azure identity assessment already built?

AzureWise Identity Security Assessment turns Microsoft Entra ID telemetry into a weighted Security Rating, detailed findings, and prioritized remediation guidance inside Azure Monitor.

Get AzureWise — $99