Elastic Adds Azure AD Graph Activity Log Ingestion and Detection Guidance

Summary
Elastic's Azure integration now ingests Azure AD Graph Activity Logs, enabling defenders to hunt enumeration using user-agent, API-version, client/user, ASN and 4xx patterns. The article also provides setup steps, detection rules and mitigations.
Key points
- Azure AD Graph Activity Logs became available in early 2026, after years when defenders lacked customer-accessible logs for this legacy API; it remains queryable in many tenants.
- Elastic routes the logs through its Azure integration into a dedicated data stream with ECS fields for actions, status codes, users, IPs and user agents.
- The article explains how to enable the diagnostic category in Entra ID, forward events through an event hub and verify that data is arriving.
- In a test tenant, ROADrecon generated just over 2,000 AAD Graph calls in roughly one minute while enumerating directory objects.
- Detection guidance covers suspicious user agents, high 4xx ratios, unusual client-and-user pairs, unusual ASN origins, ROADrecon-like request bursts and device-code sign-ins followed by enumeration.
- Suggested mitigations include auditing applications that use AAD Graph, setting the per-app blockAzureADGraphAccess control where appropriate, applying Conditional Access and disabling device-code flow for users who do not need it.
Article Details
- Defense Focus
- Ingest Azure AD Graph Activity Logs and detect directory enumeration, suspicious client and user activity, and abuse of legacy API access.
- Detection Methods
- Baseline legitimate callers and flag unfamiliar or suspicious user-agent strings, including tooling-associated agents.
- Detect unusually high 4xx response ratios and bursts of AAD Graph requests.
- Identify new OAuth client and signed-in-user pairings, including possible FOCI swaps or stolen-token use.
- Flag AAD Graph activity from unusual ASN origins.
- Detect ROADrecon-like enumeration using an aiohttp user agent combined with 500 or more requests.
- Correlate successful device-code sign-ins to the AAD Graph audience with subsequent directory enumeration by the same user within five minutes.
- Data Sources
- AzureADGraphActivityLogs diagnostic-category events routed to a queryable destination.
- Parsed AAD Graph fields including event.action, event.outcome, http.request.method, http.response.status_code, source.ip, user.id, user_agent.original, and url.path.
- Dataset properties including app_id, api_version, actor_type, roles, scopes, client_auth_method, sign_in_activity_id, and token_issued_at.
- SignInLogs and AuditLogs for identity and event correlation.
- Rule Types
- KQL match rules
- KQL new_terms rules
- ES|QL aggregations
- EQL sequence rule
- Platforms
- SIEM/XDR platform
- Log Analytics workspace
- Event hub
- Defensive Actions
- Enable the AzureADGraphActivityLogs diagnostic setting and route events to a destination where they can be queried.
- Ingest the logs into Elastic and verify that events and expected fields are present.
- Audit applications holding tokens for graph.windows.net and set blockAzureADGraphAccess for applications that do not require AAD Graph.
- Apply Conditional Access policies targeting Azure AD Graph; account for already-valid tokens remaining usable until expiry.
- Require managed and compliant devices for relevant FOCI clients, and disable device-code flow for users who do not need it.
- For service-principal callers, consider Conditional Access controls available through Workload Identities Premium.
MITRE ATT&CK
T1069 · Permission Groups DiscoveryROADrecon gathers directory roles, role assignments, and eligible role assignments to enumerate privileged access.T1078 · Valid AccountsThe described device-code sign-in and subsequent AAD Graph access use valid user credentials or tokens.T1087 · Account DiscoveryROADrecon gather enumerates directory objects including users, groups, service principals, applications, and devices.T1110 · Brute ForceThe article identifies bursts of 403 and 404 responses from endpoint or object-ID probing as a possible brute-force signal.T1526 · Cloud Service DiscoveryROADrecon and related activity make bulk calls to AAD Graph to discover Azure AD/Entra ID objects and services.T1528 · Steal Application Access TokenThe article describes phished refresh tokens and stolen tokens being used under unfamiliar clients to access AAD Graph.T1580 · Cloud Infrastructure DiscoveryThe described enumeration collects tenant details, applications, service principals, and policies as cloud infrastructure information.
People
Vendors
ElasticElastic now ingests Azure AD Graph Activity Logs into SIEM/XDR, closing a long-standing visibility gap that left AAD Graph enumeration and related adversary tooling largely unseen.MicrosoftMicrosoft Graph Activity Logs went GA in April 2024, while AzureADGraphActivityLogs only became available in early 2026, leaving years of blind spots before then.
Products
Azure AD GraphElastic now ingests Azure AD Graph Activity Logs into SIEM/XDR, closing a long-standing visibility gap that left AAD Graph enumeration and related adversary tooling largely unseen.Elastic SecurityAbout Elastic Security LabsEntra IDThe article explains how to enable the Azure integration, configure Entra ID diagnostic settings, and validate that AAD Graph events are flowing into Elastic.Microsoft GraphMicrosoft Graph Activity Logs went GA in April 2024, while AzureADGraphActivityLogs only became available in early 2026, leaving years of blind spots before then.
Tools
AADInternals#AADGraphActivityLogsAzureHound[File / tool names ] Offensive or research tooling associated with AAD Graph activity – roadrecon, ROADtools, AADInternals, MSOLSpray, Microburst, AzureHound, BloodHoundBloodHound[File / tool names ] Offensive or research tooling associated with AAD Graph activity – roadrecon, ROADtools, AADInternals, MSOLSpray, Microburst, AzureHound, BloodHoundMicroburstAdversary and research tools such as ROADtools, ROADrecon, AADInternals, MSOLSpray, and Microburst used AAD Graph heavily without producing customer-visible telemetry.MSOLSprayAdversary and research tools such as ROADtools, ROADrecon, AADInternals, MSOLSpray, and Microburst used AAD Graph heavily without producing customer-visible telemetry.ROADReconThe article shows how to enable the logs, generate and analyze traffic with ROADrecon, and detect abuse patterns such as suspicious user agents, internal API misuse, FOCI client mismatches, and 4xx spikes.ROADTools#AADGraphActivityLogs
