Detection rules › Kusto

Vaikora - Behavioral anomaly detected

This is a third-party alert feed, not a detection over modeled telemetry. The vendor product raised the finding; this rule forwards it into the SIEM. It is searchable for reference but is excluded from the detection-rule browse and the ATT&CK coverage matrix.

Status: available
Severity: medium
Time window: 1h
Group by: AgentId, Severity
Source: github.com/Azure/Azure-Sentinel

Identifies AI agent behavioral anomalies flagged by Vaikora with an anomaly score of 0.7 or above, indicating significant deviation from the agent's established behavioral baseline.

MITRE ATT&CK coverage

Tactic	Techniques
Execution	`T1059` Command and Scripting Interpreter
Stealth	`T1027` Obfuscated Files or Information

Rule body

id: e61258ec-1a7f-454c-95b5-458a6edb1ea4
name: Vaikora - Behavioral anomaly detected
description: |
  Identifies AI agent behavioral anomalies flagged by Vaikora with an anomaly score of 0.7 or above, indicating significant deviation from the agent's established behavioral baseline.
severity: Medium
status: Available
requiredDataConnectors:
  - connectorId: VaikoraSentinel
    dataTypes:
      - Vaikora_AgentSignals_CL
queryFrequency: 30m
queryPeriod: 1h
triggerOperator: GreaterThan
triggerThreshold: 0
tactics:
  - DefenseEvasion
  - Execution
relevantTechniques:
  - T1059
  - T1027
query: |
  Vaikora_AgentSignals_CL
  | where TimeGenerated > ago(1h)
  | where is_anomaly_b == true
  | where anomaly_score_d >= 0.7
  | summarize
      AnomalyCount = count(),
      MaxAnomalyScore = max(anomaly_score_d),
      AvgAnomalyScore = avg(anomaly_score_d),
      AnomalyReasons = make_set(anomaly_reason_s),
      ActionTypes = make_set(action_type_s)
    by AgentId = agent_id_s, Severity = severity_s
  | extend
      ReasonList = strcat_array(AnomalyReasons, "; "),
      ActionList = strcat_array(ActionTypes, ", ")
suppressionDuration: 30m
suppressionEnabled: false
incidentConfiguration:
  createIncident: true
  groupingConfiguration:
    enabled: true
    reopenClosedIncident: false
    lookbackDuration: 1h
    matchingMethod: Selected
    groupByEntities:
      - Account
eventGroupingSettings:
  aggregationKind: AlertPerResult
customDetails:
  MaxAnomalyScore: MaxAnomalyScore
  AvgAnomalyScore: AvgAnomalyScore
  AnomalyCount: AnomalyCount
  AnomalyReasons: ReasonList
  ActionTypes: ActionList
entityMappings:
  - entityType: Account
    fieldMappings:
      - identifier: Name
        columnName: AgentId
version: 1.0.0
kind: Scheduled

Stages and Predicates

Stage 1: `source`

Vaikora_AgentSignals_CL

Stage 2: `where`

| where TimeGenerated > ago(1h)

Stage 3: `where`

| where is_anomaly_b == true

Stage 4: `where`

| where anomaly_score_d >= 0.7

Stage 5: `summarize`

| summarize
    AnomalyCount = count(),
    MaxAnomalyScore = max(anomaly_score_d),
    AvgAnomalyScore = avg(anomaly_score_d),
    AnomalyReasons = make_set(anomaly_reason_s),
    ActionTypes = make_set(action_type_s)
  by AgentId = agent_id_s, Severity = severity_s

Stage 6: `extend`

| extend
    ReasonList = strcat_array(AnomalyReasons, "; "),
    ActionList = strcat_array(ActionTypes, ", ")

Indicators

These rows show field, operator, and value matches.

Field	Kind	Values	Search
`anomaly_score_d`	ge	`0.7`	`field:"anomaly_score_d" kind:ge value:"0.7"`
`is_anomaly_b`	eq	`true`	`field:"is_anomaly_b" kind:eq value:"true"`

Output fields

These fields are emitted when the rule matches.

Field	Source
`ActionTypes`	`summarize`
`AgentId`	`summarize`
`AnomalyCount`	`summarize`
`AnomalyReasons`	`summarize`
`AvgAnomalyScore`	`summarize`
`MaxAnomalyScore`	`summarize`
`Severity`	`summarize`
`ActionList`	`extend`
`ReasonList`	`extend`

Vaikora - Behavioral anomaly detected

MITRE ATT&CK coverage

Rule body

Stages and Predicates

Stage 1: source

Stage 2: where

Stage 3: where

Stage 4: where

Stage 5: summarize

Stage 6: extend