Vaikora - High severity AI agent action detected

Back


Id	15c49777-7cb7-4746-8064-6fa4c7a73df8
Rulename	Vaikora - High severity AI agent action detected
Description	Identifies AI agent actions from Vaikora classified as high or critical severity. These events may indicate an agent operating outside safe parameters or triggering policy thresholds.
Severity	High
Tactics	Impact Execution PrivilegeEscalation
Techniques	T1059 T1078 T1548
Required data connectors	VaikoraSentinel
Kind	Scheduled
Query frequency	1h
Query period	1h
Trigger threshold	0
Trigger operator	GreaterThan
Source Uri	https://github.com/Azure/Azure-Sentinel/blob/master/Solutions/Vaikora-Sentinel/Analytic Rules/Vaikora - High Risk AI Agent Action.yaml
Version	1.0.0
Arm template	15c49777-7cb7-4746-8064-6fa4c7a73df8.json

KQL

Vaikora_AgentSignals_CL
| where TimeGenerated > ago(1h)
| where severity_s in ("high", "critical")
| summarize
    ActionCount = count(),
    MaxAnomalyScore = max(anomaly_score_d),
    Actions = make_set(action_type_s),
    PolicyDecisions = make_set(policy_decision_s),
    ResourceTypes = make_set(resource_type_s),
    LogHashes = make_set(log_hash_s)
  by AgentId = agent_id_s, Severity = severity_s
| extend
    ActionList = strcat_array(Actions, ", "),
    PolicyList = strcat_array(PolicyDecisions, ", "),
    ResourceList = strcat_array(ResourceTypes, ", ")

YAML

entityMappings:
- entityType: Account
  fieldMappings:
  - identifier: Name
    columnName: AgentId
tactics:
- Impact
- Execution
- PrivilegeEscalation
suppressionEnabled: false
suppressionDuration: 1h
requiredDataConnectors:
- dataTypes:
  - Vaikora_AgentSignals_CL
  connectorId: VaikoraSentinel
incidentConfiguration:
  groupingConfiguration:
    reopenClosedIncident: false
    lookbackDuration: 1h
    groupByEntities:
    - Account
    enabled: true
    matchingMethod: Selected
  createIncident: true
id: 15c49777-7cb7-4746-8064-6fa4c7a73df8
severity: High
eventGroupingSettings:
  aggregationKind: AlertPerResult
status: Available
customDetails:
  PolicyDecisions: PolicyList
  ActionCount: ActionCount
  Actions: ActionList
  ResourceTypes: ResourceList
  MaxAnomalyScore: MaxAnomalyScore
query: |
  Vaikora_AgentSignals_CL
  | where TimeGenerated > ago(1h)
  | where severity_s in ("high", "critical")
  | summarize
      ActionCount = count(),
      MaxAnomalyScore = max(anomaly_score_d),
      Actions = make_set(action_type_s),
      PolicyDecisions = make_set(policy_decision_s),
      ResourceTypes = make_set(resource_type_s),
      LogHashes = make_set(log_hash_s)
    by AgentId = agent_id_s, Severity = severity_s
  | extend
      ActionList = strcat_array(Actions, ", "),
      PolicyList = strcat_array(PolicyDecisions, ", "),
      ResourceList = strcat_array(ResourceTypes, ", ")  
OriginalUri: https://github.com/Azure/Azure-Sentinel/blob/master/Solutions/Vaikora-Sentinel/Analytic Rules/Vaikora - High Risk AI Agent Action.yaml
kind: Scheduled
queryPeriod: 1h
version: 1.0.0
name: Vaikora - High severity AI agent action detected
queryFrequency: 1h
triggerThreshold: 0
relevantTechniques:
- T1059
- T1078
- T1548
description: |
    Identifies AI agent actions from Vaikora classified as high or critical severity. These events may indicate an agent operating outside safe parameters or triggering policy thresholds.
triggerOperator: GreaterThan

ARM