Before you run
Data & scope
- Required tables
- Time window
- 7 days
- Microsoft Entra ID connector with SigninLogs collected for seven days.
- Microsoft Entra ID P1 or P2 for sign-in ingestion and permission to investigate authentication details.
Documentation review by Codex (AI). No live-tenant execution or detection-performance claim. Validate schema, data coverage and expected results in your workspace.
// Hunting starter. Documentation checked 2026-11-13; not tenant-tested.
SigninLogs
| where TimeGenerated >= ago(7d) and TimeGenerated <= now()
| where IsInteractive == true and ResultType == "0"
| extend ReportedRequirement = iff(isempty(AuthenticationRequirement),
"Not reported", AuthenticationRequirement)
| summarize SignInRows = count(), Users = dcount(UserId),
FirstSeen = min(TimeGenerated), LastSeen = max(TimeGenerated),
SignInSamples = make_set(Id, 10)
by AADTenantId, AppId, AppDisplayName, ClientAppUsed,
ReportedRequirement, ConditionalAccessStatus
| order by SignInRows desc
| take 200
Read-only query. Run in your own authorized workspace after checking prerequisites.
Use this inventory to understand the reported authentication requirements associated with successful interactive access. It supports a control review: identify applications and client populations worth sampling, then inspect the evidence behind their requirements. It is not a percentage-of-users-protected-by-MFA report.
Read requirements without overclaiming
The grouping retains application ID, client type, the reported authentication requirement and aggregate Conditional Access status. Missing requirements are explicitly labelled Not reported. The query does not turn a blank value into single-factor authentication or infer a bypass from it.
Authentication requirements can be satisfied through existing token claims rather than a new prompt. Microsoft also documents that authentication details can be incomplete while events are being aggregated. Consult the original authentication and policy details before deciding what a successful event proves. A singleFactorAuthentication label is an investigation starting point, not enough evidence to conclude that every previous factor in the session was absent.
Users is an approximate distinct count within each group. The same person can appear in several applications, client types and requirement groups. Do not sum this column into a tenant-wide user denominator. SignInRows counts collected records, not business sessions or protected API operations.
Investigate the exception population
Start with sensitive administrative applications and known AI production operators. Open representative sign-ins and compare their authentication sequence with the intended policy, including exclusions, previously satisfied claims and the resource targeted by the request. Check a normal approved path before changing policy based on a suspicious-looking aggregate.
Create a review record for each unexplained population: the expected requirement, observed event IDs, owner and next validation step. Legacy clients, special-purpose accounts and access through another tenant may require different investigation paths. An exception should identify a precise business dependency and a review date.
Validate coverage
Use one known freshly prompted MFA event, one token-claim-satisfied event and one permitted single-factor event. Confirm that the inventory labels match the originals without equating labels with prompts. Check that missing data remains visible.
The hunt intentionally excludes failed and noninteractive events; use separate queries for those populations. Confirm retention and connector health before comparing week-to-week totals. This documentation-checked query has not been executed in a tenant. Its 200-group display limit and bounded samples support manual review, not an assurance claim that every identity and resource has the intended protection.
Verify the source
Microsoft references
Measure normal activity, validate known test cases, account for ingestion delays, choose a threshold and entity mapping, and assign an owner. Results require investigation.
Follow the detection learning path →