Entra ID OAuth Flow by Microsoft Authentication Broker to Device Registration Service (DRS)

Identifies separate OAuth authorization flows in Microsoft Entra ID where the same user principal and session ID are observed across multiple IP addresses within a 5-minute window. These flows involve the Microsoft Authentication Broker (MAB) as the client application and the Device Registration Service (DRS) as the target resource. This pattern is highly indicative of OAuth phishing activity, where an adversary crafts a legitimate Microsoft login URL to trick a user into completing authentication and sharing the resulting authorization code, which is then exchanged for an access and refresh token by the attacker.

Elastic rule (View on GitHub)

  1[metadata]
  2creation_date = "2025/04/30"
  3integration = ["azure"]
  4maturity = "production"
  5updated_date = "2025/12/10"
  6
  7[rule]
  8author = ["Elastic"]
  9description = """
 10Identifies separate OAuth authorization flows in Microsoft Entra ID where the same user principal and session ID are
 11observed across multiple IP addresses within a 5-minute window. These flows involve the Microsoft Authentication Broker
 12(MAB) as the client application and the Device Registration Service (DRS) as the target resource. This pattern is highly
 13indicative of OAuth phishing activity, where an adversary crafts a legitimate Microsoft login URL to trick a user into
 14completing authentication and sharing the resulting authorization code, which is then exchanged for an access and
 15refresh token by the attacker.
 16"""
 17false_positives = [
 18    """
 19    Legitimate device registrations using Microsoft Authentication Broker may occur during corporate enrollment
 20    scenarios or bulk provisioning, but it is uncommon for multiple source IPs to register the same identity across
 21    Microsoft Graph, Device Registration Service (DRS), and Azure Active Directory (AAD) in a short time span.
 22    """,
 23]
 24from = "now-61m"
 25interval = "60m"
 26language = "esql"
 27license = "Elastic License v2"
 28name = "Entra ID OAuth Flow by Microsoft Authentication Broker to Device Registration Service (DRS)"
 29note = """## Triage and analysis
 30
 31### Investigating Entra ID OAuth Flow by Microsoft Authentication Broker to Device Registration Service (DRS)
 32
 33This rule identifies potential OAuth phishing behavior in Microsoft Entra ID where two OAuth authorization flows are observed in quick succession, sharing the same user principal and session ID but originating from different IP addresses. The client application is the Microsoft Authentication Broker, and the target resource is the Device Registration Service (DRS). This pattern is indicative of adversaries attempting to phish targets for OAuth sessions by tricking users into authenticating through a crafted URL, which then allows the attacker to obtain an authorization code and exchange it for access and refresh tokens.
 34
 35### Possible Investigation Steps:
 36
 37- `target`: The user principal name targeted by the authentication broker. Investigate whether this user has recently registered a device, signed in from new IPs, or had password resets or MFA changes.
 38- `session_id`: Used to correlate all events in the OAuth flow. All sign-ins in the alert share the same session, suggesting shared or hijacked state.
 39- `unique_token_id`: Lists tokens generated in the flow. If multiple IDs exist in the same session, this indicates token issuance from different locations.
 40- `source_ip`, `city_name`, `country_name`, `region_name`: Review the IPs and geolocations involved. A mismatch in geographic origin within minutes can signal adversary involvement.
 41- `user_agent`: Conflicting user agents (e.g., `python-requests` and `Chrome`) suggest one leg of the session was scripted or automated.
 42- `os`: If multiple operating systems are observed in the same short session (e.g., macOS and Windows), this may suggest activity from different environments.
 43- `incoming_token_type`: Look for values like `"none"` or `"refreshToken"` that can indicate abnormal or re-authenticated activity.
 44- `token_session_status`: A value of `"unbound"` means the issued token is not tied to a device or CAE session, making it reusable from another IP.
 45- `conditional_access_status`: If this is `"notApplied"`, it may indicate that expected access policies were not enforced.
 46- `auth_count`: Number of events in the session. More than one indicates the session was reused within the time window.
 47- `target_time_window`: Use this to pivot into raw sign-in logs to review the exact sequence and timing of the activity.
 48- Search `azure.auditlogs` for any device join or registration activity around the `target_time_window`.
 49- Review `azure.identityprotection` logs for anonymized IPs, impossible travel, or token replay alerts.
 50- Search for other activity from the same IPs across all users to identify horizontal movement.
 51
 52### False Positive Analysis
 53
 54- A legitimate device join from a user switching networks (e.g., mobile hotspot to Wi-Fi) could explain multi-IP usage.
 55- Some identity management agents or EDR tools may use MAB for background device registration flows.
 56- Developers or IT administrators may access DRS across environments when testing.
 57
 58### Response and Remediation
 59
 60- If confirmed unauthorized, revoke all refresh tokens for the user and disable any suspicious registered devices.
 61- Notify the user and verify if the authentication or device join was expected.
 62- Review Conditional Access policies for the Microsoft Authentication Broker (`29d9ed98-a469-4536-ade2-f981bc1d605e`) to ensure enforcement of MFA and device trust.
 63- Consider restricting token-based reauthentication from anonymized infrastructure or unusual user agents.
 64- Continue monitoring for follow-on activity, such as privilege escalation, token misuse, or lateral movement.
 65"""
 66references = [
 67    "https://www.volexity.com/blog/2025/04/22/phishing-for-codes-russian-threat-actors-target-microsoft-365-oauth-workflows/",
 68    "https://github.com/dirkjanm/ROADtools",
 69    "https://dirkjanm.io/phishing-for-microsoft-entra-primary-refresh-tokens/",
 70]
 71risk_score = 73
 72rule_id = "375132c6-25d5-11f0-8745-f661ea17fbcd"
 73setup = """#### Required Microsoft Entra ID Sign-In Logs
 74This rule requires the Microsoft Entra ID Sign-In Logs integration be enabled and configured to collect sign-in logs. In Entra ID, sign-in logs must be enabled and streaming to the Event Hub used for the Azure integration.
 75"""
 76severity = "high"
 77tags = [
 78    "Domain: Cloud",
 79    "Domain: Identity",
 80    "Data Source: Azure",
 81    "Data Source: Entra ID",
 82    "Data Source: Entra ID Sign-in Logs",
 83    "Use Case: Identity and Access Audit",
 84    "Use Case: Threat Detection",
 85    "Resources: Investigation Guide",
 86    "Tactic: Initial Access",
 87]
 88timestamp_override = "event.ingested"
 89type = "esql"
 90
 91query = '''
 92from logs-azure.signinlogs-* metadata _id, _version, _index
 93| where
 94    event.dataset == "azure.signinlogs" and
 95    event.outcome == "success" and
 96    azure.signinlogs.properties.user_type == "Member" and
 97    azure.signinlogs.identity is not null and
 98    azure.signinlogs.properties.user_principal_name is not null and
 99    source.address is not null and
100    azure.signinlogs.properties.app_id == "29d9ed98-a469-4536-ade2-f981bc1d605e" and  // MAB
101    azure.signinlogs.properties.resource_id == "01cb2876-7ebd-4aa4-9cc9-d28bd4d359a9"  // DRS
102
103| eval
104    Esql.time_window_date_trunc = date_trunc(30 minutes, @timestamp),
105    Esql.azure_signinlogs_properties_session_id = azure.signinlogs.properties.session_id,
106    Esql.is_browser_case = case(
107        to_lower(azure.signinlogs.properties.device_detail.browser) rlike "(chrome|firefox|edge|safari).*", 1, 0
108    )
109
110| stats
111    Esql_priv.azure_signinlogs_properties_user_display_name_values = values(azure.signinlogs.properties.user_display_name),
112    Esql_priv.azure_signinlogs_properties_user_principal_name_values = values(azure.signinlogs.properties.user_principal_name),
113    Esql.azure_signinlogs_properties_session_id_values = values(azure.signinlogs.properties.session_id),
114    Esql.azure_signinlogs_properties_unique_token_identifier_values = values(azure.signinlogs.properties.unique_token_identifier),
115
116    Esql.source_geo_city_name_values = values(source.geo.city_name),
117    Esql.source_geo_country_name_values = values(source.geo.country_name),
118    Esql.source_geo_region_name_values = values(source.geo.region_name),
119    Esql.source_address_values = values(source.address),
120    Esql.source_address_count_distinct = count_distinct(source.address),
121    Esql.source_as_organization_name_values = values(source.`as`.organization.name),
122
123    Esql.azure_signinlogs_properties_authentication_protocol_values = values(azure.signinlogs.properties.authentication_protocol),
124    Esql.azure_signinlogs_properties_authentication_requirement_values = values(azure.signinlogs.properties.authentication_requirement),
125    Esql.azure_signinlogs_properties_is_interactive_values = values(azure.signinlogs.properties.is_interactive),
126
127    Esql.azure_signinlogs_properties_incoming_token_type_values = values(azure.signinlogs.properties.incoming_token_type),
128    Esql.azure_signinlogs_properties_token_protection_status_details_sign_in_session_status_values = values(azure.signinlogs.properties.token_protection_status_details.sign_in_session_status),
129    Esql.azure_signinlogs_properties_session_id_count_distinct = count_distinct(azure.signinlogs.properties.session_id),
130    Esql.azure_signinlogs_properties_app_display_name_values = values(azure.signinlogs.properties.app_display_name),
131    Esql.azure_signinlogs_properties_app_id_values = values(azure.signinlogs.properties.app_id),
132    Esql.azure_signinlogs_properties_resource_id_values = values(azure.signinlogs.properties.resource_id),
133    Esql.azure_signinlogs_properties_resource_display_name_values = values(azure.signinlogs.properties.resource_display_name),
134
135    Esql.azure_signinlogs_properties_app_owner_tenant_id_values = values(azure.signinlogs.properties.app_owner_tenant_id),
136    Esql.azure_signinlogs_properties_resource_owner_tenant_id_values = values(azure.signinlogs.properties.resource_owner_tenant_id),
137
138    Esql.azure_signinlogs_properties_conditional_access_status_values = values(azure.signinlogs.properties.conditional_access_status),
139    Esql.azure_signinlogs_properties_risk_state_values = values(azure.signinlogs.properties.risk_state),
140    Esql.azure_signinlogs_properties_risk_level_aggregated_values = values(azure.signinlogs.properties.risk_level_aggregated),
141
142    Esql.azure_signinlogs_properties_device_detail_browser_values = values(azure.signinlogs.properties.device_detail.browser),
143    Esql.azure_signinlogs_properties_device_detail_operating_system_values = values(azure.signinlogs.properties.device_detail.operating_system),
144    Esql.user_agent_original_values = values(user_agent.original),
145    Esql.is_browser_case_max = max(Esql.is_browser_case),
146
147    Esql.event_count = count(*)
148  by
149    Esql.time_window_date_trunc,
150    azure.signinlogs.properties.user_principal_name,
151    azure.signinlogs.properties.session_id
152
153| keep
154    Esql.time_window_date_trunc,
155    Esql_priv.azure_signinlogs_properties_user_display_name_values,
156    Esql_priv.azure_signinlogs_properties_user_principal_name_values,
157    Esql.azure_signinlogs_properties_session_id_values,
158    Esql.azure_signinlogs_properties_unique_token_identifier_values,
159    Esql.source_geo_city_name_values,
160    Esql.source_geo_country_name_values,
161    Esql.source_geo_region_name_values,
162    Esql.source_address_values,
163    Esql.source_address_count_distinct,
164    Esql.source_as_organization_name_values,
165    Esql.azure_signinlogs_properties_authentication_protocol_values,
166    Esql.azure_signinlogs_properties_authentication_requirement_values,
167    Esql.azure_signinlogs_properties_is_interactive_values,
168    Esql.azure_signinlogs_properties_incoming_token_type_values,
169    Esql.azure_signinlogs_properties_token_protection_status_details_sign_in_session_status_values,
170    Esql.azure_signinlogs_properties_session_id_count_distinct,
171    Esql.azure_signinlogs_properties_app_display_name_values,
172    Esql.azure_signinlogs_properties_app_id_values,
173    Esql.azure_signinlogs_properties_resource_id_values,
174    Esql.azure_signinlogs_properties_resource_display_name_values,
175    Esql.azure_signinlogs_properties_app_owner_tenant_id_values,
176    Esql.azure_signinlogs_properties_resource_owner_tenant_id_values,
177    Esql.azure_signinlogs_properties_conditional_access_status_values,
178    Esql.azure_signinlogs_properties_risk_state_values,
179    Esql.azure_signinlogs_properties_risk_level_aggregated_values,
180    Esql.azure_signinlogs_properties_device_detail_browser_values,
181    Esql.azure_signinlogs_properties_device_detail_operating_system_values,
182    Esql.user_agent_original_values,
183    Esql.is_browser_case_max,
184    Esql.event_count
185
186| where
187    Esql.source_address_count_distinct >= 2 and
188    Esql.azure_signinlogs_properties_session_id_count_distinct == 1 and
189    Esql.is_browser_case_max >= 1 and
190    Esql.event_count >= 2
191'''
192
193
194[[rule.threat]]
195framework = "MITRE ATT&CK"
196[[rule.threat.technique]]
197id = "T1078"
198name = "Valid Accounts"
199reference = "https://attack.mitre.org/techniques/T1078/"
200[[rule.threat.technique.subtechnique]]
201id = "T1078.004"
202name = "Cloud Accounts"
203reference = "https://attack.mitre.org/techniques/T1078/004/"
204
205[[rule.threat.technique]]
206id = "T1566"
207name = "Phishing"
208reference = "https://attack.mitre.org/techniques/T1566/"
209[[rule.threat.technique.subtechnique]]
210id = "T1566.002"
211name = "Spearphishing Link"
212reference = "https://attack.mitre.org/techniques/T1566/002/"
213
214
215
216[rule.threat.tactic]
217id = "TA0001"
218name = "Initial Access"
219reference = "https://attack.mitre.org/tactics/TA0001/"
220[[rule.threat]]
221framework = "MITRE ATT&CK"
222[[rule.threat.technique]]
223id = "T1528"
224name = "Steal Application Access Token"
225reference = "https://attack.mitre.org/techniques/T1528/"
226
227
228
229[rule.threat.tactic]
230id = "TA0006"
231name = "Credential Access"
232reference = "https://attack.mitre.org/tactics/TA0006/"

Triage and analysis

Investigating Entra ID OAuth Flow by Microsoft Authentication Broker to Device Registration Service (DRS)

This rule identifies potential OAuth phishing behavior in Microsoft Entra ID where two OAuth authorization flows are observed in quick succession, sharing the same user principal and session ID but originating from different IP addresses. The client application is the Microsoft Authentication Broker, and the target resource is the Device Registration Service (DRS). This pattern is indicative of adversaries attempting to phish targets for OAuth sessions by tricking users into authenticating through a crafted URL, which then allows the attacker to obtain an authorization code and exchange it for access and refresh tokens.

Possible Investigation Steps:

  • target: The user principal name targeted by the authentication broker. Investigate whether this user has recently registered a device, signed in from new IPs, or had password resets or MFA changes.
  • session_id: Used to correlate all events in the OAuth flow. All sign-ins in the alert share the same session, suggesting shared or hijacked state.
  • unique_token_id: Lists tokens generated in the flow. If multiple IDs exist in the same session, this indicates token issuance from different locations.
  • source_ip, city_name, country_name, region_name: Review the IPs and geolocations involved. A mismatch in geographic origin within minutes can signal adversary involvement.
  • user_agent: Conflicting user agents (e.g., python-requests and Chrome) suggest one leg of the session was scripted or automated.
  • os: If multiple operating systems are observed in the same short session (e.g., macOS and Windows), this may suggest activity from different environments.
  • incoming_token_type: Look for values like "none" or "refreshToken" that can indicate abnormal or re-authenticated activity.
  • token_session_status: A value of "unbound" means the issued token is not tied to a device or CAE session, making it reusable from another IP.
  • conditional_access_status: If this is "notApplied", it may indicate that expected access policies were not enforced.
  • auth_count: Number of events in the session. More than one indicates the session was reused within the time window.
  • target_time_window: Use this to pivot into raw sign-in logs to review the exact sequence and timing of the activity.
  • Search azure.auditlogs for any device join or registration activity around the target_time_window.
  • Review azure.identityprotection logs for anonymized IPs, impossible travel, or token replay alerts.
  • Search for other activity from the same IPs across all users to identify horizontal movement.

False Positive Analysis

  • A legitimate device join from a user switching networks (e.g., mobile hotspot to Wi-Fi) could explain multi-IP usage.
  • Some identity management agents or EDR tools may use MAB for background device registration flows.
  • Developers or IT administrators may access DRS across environments when testing.

Response and Remediation

  • If confirmed unauthorized, revoke all refresh tokens for the user and disable any suspicious registered devices.
  • Notify the user and verify if the authentication or device join was expected.
  • Review Conditional Access policies for the Microsoft Authentication Broker (29d9ed98-a469-4536-ade2-f981bc1d605e) to ensure enforcement of MFA and device trust.
  • Consider restricting token-based reauthentication from anonymized infrastructure or unusual user agents.
  • Continue monitoring for follow-on activity, such as privilege escalation, token misuse, or lateral movement.

References

Related rules

to-top