detection.wiki

Detection rules › Kusto

AIShield - Image classification AI Model Evasion high suspicious vulnerability detection

Status
available
Severity
high
Time window
1h
Source
github.com/Azure/Azure-Sentinel

'This alert creates an incident when Image classification AI Model Evasion high suspicious, high severity vulnerability detected from the AIShield.'

Rule body kusto

id: 76c9e83d-c6f9-4270-ad21-761410f9cc6c
name: AIShield - Image classification AI Model Evasion high suspicious vulnerability detection
description: |
  'This alert creates an incident when Image classification AI Model Evasion high suspicious, high severity vulnerability detected from the AIShield.'
severity: High
status: Available
requiredDataConnectors:
  - connectorId: BoschAIShield
    dataTypes:
      - AIShield
queryFrequency: 1h
queryPeriod: 1h
triggerOperator: gt
triggerThreshold: 0
tactics: []
relevantTechniques: []
query: |
  AIShield
  | where Message has 'Image Classification AI Model Evasion Attack Identified'
  | where Severity =~ 'High'
  | where SuspiciousLevel =~ 'high suspicious attack'
  | extend NTDomain = tostring(split(Computer, '\\')[0]), HostName = tostring(split(Computer, '.')[0])
entityMappings:
  - entityType: Host
    fieldMappings:
      - identifier: HostName
        columnName: HostName
      - identifier: NTDomain
        columnName: NTDomain
  - entityType: IP
    fieldMappings:
      - identifier: Address
        columnName: SourceName
eventGroupingSettings:
  aggregationKind: SingleAlert
alertDetailsOverride:
  alertDisplayNameFormat: AIShield - Image Classification AI Model Evasion high suspicious vulnerability detected.
  alertDescriptionFormat: |
    This query detects Image Classification AI Model Evasion high suspicious, high severity alert from AIShield generated at {{TimeGenerated}}.\n\nPlease check the source for more information and investigate further.
  alertTacticsColumnName: null
  alertSeverityColumnName: Severity
version: 1.0.4
kind: Scheduled

Stages and Predicates

Stage 1: source

AIShield

Stage 2: where

| where Message has 'Image Classification AI Model Evasion Attack Identified'

Stage 3: where

| where Severity =~ 'High'

Stage 4: where

| where SuspiciousLevel =~ 'high suspicious attack'

Stage 5: extend

| extend NTDomain = tostring(split(Computer, '\\')[0]), HostName = tostring(split(Computer, '.')[0])

Indicators

Each row is a field, operator, and value that the rule matches. The corpus column counts how many other rules in the catalog look for the same combination: high numbers point to widely-used, community-vetted indicators. Blank or 1 shows that the indicator is specific to this rule.

FieldKindValues
Messagematch
  • Image Classification AI Model Evasion Attack Identified transforms: term
Severityeq
  • High
SuspiciousLeveleq
  • high suspicious attack

Output fields

Fields the rule emits when it matches. Chronicle authors list these in the outcome block; they appear on the detection and $risk_score drives alerting. Sentinel / Defender XDR rules build them up through project / summarize / extend stages. Sentinel maps these into alert fields via entityMappings and customDetails; Defender XDR custom detections surface them as alert fields directly.

FieldSource
HostNameextend
NTDomainextend