Detection rules › Kusto

SAP BTP - Unaudited custom app with login-only activity

Status
available
Severity
medium
Time window
7d
Group by
AlsServiceId, ClientId, Tenant, XsAppName
Source
github.com/Azure/Azure-Sentinel

Identifies SAP BTP custom applications (CloudFoundry, SAP CAP, etc.) that only produce XSUAA authentication events (TokenIssuedEvent, ClientAuthenticationSuccess) but have not generated any business audit log activity in the past 7 days. This pattern indicates that the application has not implemented audit logging (e.g., missing @AuditLog annotations in CAP or missing audit log service bindings), creating a security blind spot where user actions within the application are invisible to monitoring. The 7-day lookback avoids false positives for properly instrumented apps whose users simply have not performed auditable actions in the current session. Attackers could exploit such unaudited applications to perform malicious operations without detection.

MITRE ATT&CK coverage

Rule body

id: 5e8f2a1b-7c3d-4b9e-a6f0-1d2e3c4b5a6f
kind: Scheduled
name: SAP BTP - Unaudited custom app with login-only activity
description: |
  Identifies SAP BTP custom applications (CloudFoundry, SAP CAP, etc.) that only produce
  XSUAA authentication events (TokenIssuedEvent, ClientAuthenticationSuccess) but have not
  generated any business audit log activity in the past 7 days. This pattern indicates that
  the application has not implemented audit logging (e.g., missing @AuditLog annotations in
  CAP or missing audit log service bindings), creating a security blind spot where user
  actions within the application are invisible to monitoring. The 7-day lookback avoids
  false positives for properly instrumented apps whose users simply have not performed
  auditable actions in the current session. Attackers could exploit such unaudited
  applications to perform malicious operations without detection.
severity: Medium
status: Available
requiredDataConnectors:
  - connectorId: SAPBTPAuditEvents
    dataTypes:
      - SAPBTPAuditLog_CL
queryFrequency: 1h
queryPeriod: 7d
triggerOperator: gt
triggerThreshold: 0
tactics:
  - DefenseEvasion
relevantTechniques:
  - T1562
  - T1562.008
query: |
  // Lookback period for audit events - a longer window avoids false positives for
  // properly instrumented apps where users logged in but haven't performed actions yet
  let audit_lookback = ago(7d);
  // Known BTP platform service patterns (excluded from detection)
  let platform_service_patterns = dynamic([
      "app-studio", "auditlog", "cis-local", "service-manager",
      "destination-xsappname", "connectivity-proxy", "feature-flags"
  ]);
  // Step 1: From XSUAA TokenIssuedEvent, identify custom apps with interactive user logins
  // XSUAA is a shared subaccount service - the client_id in each token identifies the actual app
  let app_logins = SAPBTPAuditLog_CL
      // Scope logins to the current 1h run cycle (matches queryFrequency)
      | where TimeGenerated > ago(1h)
      | where Category == "audit.security-events"
      | extend data_s = tostring(Message.data)
      | where data_s has "TokenIssuedEvent"
      | extend ParsedData = parse_json(data_s)
      // origin field = client_id of the app the token was issued for
      | extend ClientId = tostring(ParsedData.origin)
      | where ClientId startswith "sb-"
      // Parse nested event message for human user and grant type
      | extend EventMessage = tostring(ParsedData.message)
      | extend GrantType = extract(@'"grant_type"\s*:\s*"([^"]+)"', 1, EventMessage)
      // Only interactive browser-based logins (not service-to-service tokens)
      | where GrantType == "authorization_code"
      // Derive app identifiers: sb-<XsAppName>!t<number>
      | extend XsAppName = extract(@"^sb-(.+?)!\w+$", 1, ClientId)
      | where isnotempty(XsAppName)
      // Exclude known platform services
      | where not(XsAppName has_any (platform_service_patterns))
      // Extract human-readable app name (first segment before subaccount qualifier)
      | extend AppName = extract(@"^([A-Za-z][A-Za-z0-9_]*)", 1, XsAppName)
      | extend HumanUser = extract(@'"user_name"\s*:\s*"([^"]+)"', 1, EventMessage)
      | extend IPAddress = tostring(Message.ip);
  // Step 2a: Identify apps that have business audit events where the service binding
  // identity appears in UserName (e.g. platform-generated audit events)
  let audited_apps = SAPBTPAuditLog_CL
      | where TimeGenerated > audit_lookback
      | where Category !in ("audit.security-events")
      | where UserName startswith "sb-"
      | extend AuditedAppName = extract(@"^sb-(.+?)!\w+", 1, UserName)
      | where isnotempty(AuditedAppName)
      | distinct AuditedAppName, Tenant;
  // Step 2b: Identify CF app instances that have an audit log service binding.
  // When a CAP/custom app binds the audit log service, BTP provisions a
  // 'customer-auditlog' service key whose XSUAA token events share the same
  // AlsServiceId (CF app GUID) as the app's own login tokens. This gives a
  // reliable per-app correlation that works even when the audit log service
  // instance is deployed in a different CF space than the app itself.
  let audited_app_instances = SAPBTPAuditLog_CL
      | where TimeGenerated > audit_lookback
      | where Category == "audit.security-events"
      | where UserName contains "customer-auditlog"
      | where isnotempty(AlsServiceId)
      | distinct AlsServiceId, Tenant;
  // Step 3: Find custom apps with logins but no business audit trail
  app_logins
  | join kind=leftanti (audited_apps | project XsAppName = AuditedAppName, Tenant) on XsAppName, Tenant
  | join kind=leftanti (audited_app_instances) on AlsServiceId, Tenant
  | summarize
      LoginCount = count(),
      FirstLogin = min(TimeGenerated),
      LastLogin = max(TimeGenerated),
      Users = make_set(HumanUser, 100),
      IPs = make_set(IPAddress, 100),
      AppName = take_any(AppName),
      SubaccountName = take_any(SubaccountName),
      OrgId = take_any(OrgId),
      SpaceId = take_any(SpaceId)
      by XsAppName, ClientId, Tenant
  | extend Users = set_difference(Users, dynamic([""]))
  | project
      FirstLogin,
      LastLogin,
      AppName,
      XsAppName,
      ClientId,
      SubaccountName,
      Tenant,
      OrgId,
      SpaceId,
      LoginCount,
      Users,
      IPs,
      CloudApp = "SAP BTP"
eventGroupingSettings:
  aggregationKind: AlertPerResult
entityMappings:
  - entityType: CloudApplication
    fieldMappings:
      - identifier: Name
        columnName: CloudApp
alertDetailsOverride:
  alertDisplayNameFormat: "SAP BTP: Unaudited custom app '{{AppName}}' - login-only activity detected"
  alertDescriptionFormat: |
    The custom app '{{AppName}}' ({{XsAppName}}) in subaccount '{{SubaccountName}}' produced only
    XSUAA authentication events but no business audit trail within the past 7 days.
    
    This indicates the app has not implemented audit logging, creating a security blind spot
    where user actions within the application are invisible to Microsoft Sentinel.
    
    Recommended actions:
    1. Review the application's audit log implementation
    2. Add the audit log service binding in SAP BTP cockpit
    3. For CAP apps, implement @AuditLog annotations on sensitive entities and services
    4. Investigate user activity through alternative logs (application logs, HTTP access logs)
customDetails:
  AppName: AppName
  XsAppName: XsAppName
  ClientId: ClientId
  SubaccountName: SubaccountName
  Tenant: Tenant
  OrgId: OrgId
  SpaceId: SpaceId
  LoginCount: LoginCount
  Users: Users
  IPs: IPs
version: 1.0.4

Stages and Predicates

Parameters

let audit_lookback = ago(7d);

let app_logins is inlined into the numbered stages below.

Let binding: platform_service_patterns used in Stage 14

let platform_service_patterns = dynamic([
    "app-studio", "auditlog", "cis-local", "service-manager",
    "destination-xsappname", "connectivity-proxy", "feature-flags"
]);

Let binding: audited_apps used in Stage 16

let audited_apps = SAPBTPAuditLog_CL
    | where TimeGenerated > audit_lookback
    | where Category !in ("audit.security-events")
    | where UserName startswith "sb-"
    | extend AuditedAppName = extract(@"^sb-(.+?)!\w+", 1, UserName)
    | where isnotempty(AuditedAppName)
    | distinct AuditedAppName, Tenant;

Let binding: audited_app_instances used in Stage 17

let audited_app_instances = SAPBTPAuditLog_CL
    | where TimeGenerated > audit_lookback
    | where Category == "audit.security-events"
    | where UserName contains "customer-auditlog"
    | where isnotempty(AlsServiceId)
    | distinct AlsServiceId, Tenant;

Stages 1 to 15 define let app_logins (the rule's main pipeline source); stages 16 to 20 run on it.

Stage 1: source

SAPBTPAuditLog_CL

Stage 2: where

| where TimeGenerated > ago(1h)

Stage 3: where

| where Category == "audit.security-events"

Stage 4: extend

| extend data_s = tostring(Message.data)

Stage 5: where

| where data_s has "TokenIssuedEvent"

Stage 6: extend

| extend ParsedData = parse_json(data_s)

Stage 7: extend

| extend ClientId = tostring(ParsedData.origin)

Stage 8: where

| where ClientId startswith "sb-"

Stage 9: extend

| extend EventMessage = tostring(ParsedData.message)

Stage 10: extend

| extend GrantType = extract(@'"grant_type"\s*:\s*"([^"]+)"', 1, EventMessage)

Stage 11: where

| where GrantType == "authorization_code"

Stage 12: extend

| extend XsAppName = extract(@"^sb-(.+?)!\w+$", 1, ClientId)

Stage 13: where

| where isnotempty(XsAppName)

Stage 14: where

| where not(XsAppName has_any (platform_service_patterns))

Stage 15: extend (3 consecutive steps)

| extend AppName = extract(@"^([A-Za-z][A-Za-z0-9_]*)", 1, XsAppName)
| extend HumanUser = extract(@'"user_name"\s*:\s*"([^"]+)"', 1, EventMessage)
| extend IPAddress = tostring(Message.ip)

Stage 16: join (negated)

app_logins
| join kind=leftanti (audited_apps | project XsAppName = AuditedAppName, Tenant) on XsAppName, Tenant

Stage 17: join (negated)

| join kind=leftanti (audited_app_instances) on AlsServiceId, Tenant

Stage 18: summarize

| summarize
    LoginCount = count(),
    FirstLogin = min(TimeGenerated),
    LastLogin = max(TimeGenerated),
    Users = make_set(HumanUser, 100),
    IPs = make_set(IPAddress, 100),
    AppName = take_any(AppName),
    SubaccountName = take_any(SubaccountName),
    OrgId = take_any(OrgId),
    SpaceId = take_any(SpaceId)
    by XsAppName, ClientId, Tenant

Stage 19: extend

| extend Users = set_difference(Users, dynamic([""]))

Stage 20: project

| project
    FirstLogin,
    LastLogin,
    AppName,
    XsAppName,
    ClientId,
    SubaccountName,
    Tenant,
    OrgId,
    SpaceId,
    LoginCount,
    Users,
    IPs,
    CloudApp = "SAP BTP"

Exclusions

The rule actively suppresses these predicates.

FieldKindExcluded valuesSearch
XsAppNamematchapp-studio, auditlog, cis-local, service-manager, destination-xsappname, connectivity-proxy, feature-flagsexcludes:XsAppName
AuditedAppNameis_not_null(no value, null check)excludes:AuditedAppName
TimeGeneratedcross_field_compareaudit_lookbackexcludes:TimeGenerated field:"TimeGenerated" value:"audit_lookback"
UserNamestarts_withsb-excludes:UserName field:"UserName" value:"sb-"
AlsServiceIdis_not_null(no value, null check)excludes:AlsServiceId
Categoryeqaudit.security-eventsexcludes:Category field:"Category" value:"audit.security-events"
UserNamecontainscustomer-auditlogexcludes:UserName field:"UserName" value:"customer-auditlog"

Indicators

These rows show field, operator, and value matches.

Output fields

These fields are emitted when the rule matches.

FieldSource
AppNameproject
ClientIdproject
CloudAppproject
FirstLoginproject
IPsproject
LastLoginproject
LoginCountproject
OrgIdproject
SpaceIdproject
SubaccountNameproject
Tenantproject
Usersproject
XsAppNameproject