Copilot - Jailbreak Attempt Detected | Microsoft Sentinel Analytic Rules

Back


Id	e5f6a7b8-c9d0-41e2-f3a4-b5c6d7e8f9a0
Rulename	Copilot - Jailbreak Attempt Detected
Description	Detects jailbreak attempts in Copilot interactions where users are trying to bypass Copilot guardrails and security controls. This rule identifies prompt injection and LLM abuse scenarios that could lead to initial access, credential access, or system impact.
Severity	High
Tactics	InitialAccess CredentialAccess Impact
Techniques	T1078 T1110 T1565
Required data connectors	MicrosoftCopilot
Kind	Scheduled
Query frequency	5m
Query period	5m
Trigger threshold	0
Trigger operator	gt
Source Uri	https://github.com/Azure/Azure-Sentinel/blob/master/Solutions/Microsoft%20Copilot/Analytic%20Rules/CopilotJailbreakAttempt.yaml
Version	1.0.0
Arm template	e5f6a7b8-c9d0-41e2-f3a4-b5c6d7e8f9a0.json

KQL

CopilotActivity
| where RecordType == "CopilotInteraction"
| where LLMEventData has "JailbreakDetected"
| extend Data = parse_json(LLMEventData)
| extend Jailbreak = tostring(Data.Messages[0].JailbreakDetected)
| where Jailbreak == "true"
| project TimeGenerated, ActorName, AIModelName, Jailbreak

YAML

severity: High
tactics:
- InitialAccess
- CredentialAccess
- Impact
queryFrequency: 5m
id: e5f6a7b8-c9d0-41e2-f3a4-b5c6d7e8f9a0
query: |
  CopilotActivity
  | where RecordType == "CopilotInteraction"
  | where LLMEventData has "JailbreakDetected"
  | extend Data = parse_json(LLMEventData)
  | extend Jailbreak = tostring(Data.Messages[0].JailbreakDetected)
  | where Jailbreak == "true"
  | project TimeGenerated, ActorName, AIModelName, Jailbreak
entityMappings:
- fieldMappings:
  - identifier: FullName
    columnName: ActorName
  entityType: Account
kind: Scheduled
triggerOperator: gt
name: Copilot - Jailbreak Attempt Detected
version: 1.0.0
status: Available
requiredDataConnectors:
- dataTypes:
  - CopilotActivity
  connectorId: MicrosoftCopilot
description: |
  'Detects jailbreak attempts in Copilot interactions where users are trying to bypass Copilot guardrails and security controls.
  This rule identifies prompt injection and LLM abuse scenarios that could lead to initial access, credential access, or system impact.'
relevantTechniques:
- T1078
- T1110
- T1565
queryPeriod: 5m
OriginalUri: https://github.com/Azure/Azure-Sentinel/blob/master/Solutions/Microsoft%20Copilot/Analytic%20Rules/CopilotJailbreakAttempt.yaml
triggerThreshold: 0

ARM

{
  "$schema": "https://schema.management.azure.com/schemas/2019-04-01/deploymentTemplate.json#",
  "contentVersion": "1.0.0.0",
  "parameters": {
    "workspace": {
      "type": "String"
    }
  },
  "resources": [
    {
      "apiVersion": "2024-01-01-preview",
      "id": "[concat(resourceId('Microsoft.OperationalInsights/workspaces/providers', parameters('workspace'), 'Microsoft.SecurityInsights'),'/alertRules/e5f6a7b8-c9d0-41e2-f3a4-b5c6d7e8f9a0')]",
      "kind": "Scheduled",
      "name": "[concat(parameters('workspace'),'/Microsoft.SecurityInsights/e5f6a7b8-c9d0-41e2-f3a4-b5c6d7e8f9a0')]",
      "properties": {
        "alertRuleTemplateName": "e5f6a7b8-c9d0-41e2-f3a4-b5c6d7e8f9a0",
        "customDetails": null,
        "description": "'Detects jailbreak attempts in Copilot interactions where users are trying to bypass Copilot guardrails and security controls.\nThis rule identifies prompt injection and LLM abuse scenarios that could lead to initial access, credential access, or system impact.'\n",
        "displayName": "Copilot - Jailbreak Attempt Detected",
        "enabled": true,
        "entityMappings": [
          {
            "entityType": "Account",
            "fieldMappings": [
              {
                "columnName": "ActorName",
                "identifier": "FullName"
              }
            ]
          }
        ],
        "OriginalUri": "https://github.com/Azure/Azure-Sentinel/blob/master/Solutions/Microsoft%20Copilot/Analytic%20Rules/CopilotJailbreakAttempt.yaml",
        "query": "CopilotActivity\n| where RecordType == \"CopilotInteraction\"\n| where LLMEventData has \"JailbreakDetected\"\n| extend Data = parse_json(LLMEventData)\n| extend Jailbreak = tostring(Data.Messages[0].JailbreakDetected)\n| where Jailbreak == \"true\"\n| project TimeGenerated, ActorName, AIModelName, Jailbreak\n",
        "queryFrequency": "PT5M",
        "queryPeriod": "PT5M",
        "severity": "High",
        "status": "Available",
        "subTechniques": [],
        "suppressionDuration": "PT1H",
        "suppressionEnabled": false,
        "tactics": [
          "CredentialAccess",
          "Impact",
          "InitialAccess"
        ],
        "techniques": [
          "T1078",
          "T1110",
          "T1565"
        ],
        "templateVersion": "1.0.0",
        "triggerOperator": "GreaterThan",
        "triggerThreshold": 0
      },
      "type": "Microsoft.OperationalInsights/workspaces/providers/alertRules"
    }
  ]
}