Skip to main content

Microsoft Sentinel Integration

Forward Noxys AI interaction events and security alerts to Microsoft Sentinel for centralized security monitoring and threat detection.

Overview​

This integration enables:

  • Real-time event streaming from Noxys to Sentinel via webhook
  • Custom analytics rules to detect AI-related threats
  • Automated response playbooks for policy violations
  • Unified dashboard combining Noxys and Sentinel data

Data Flow:

Noxys (Policy Violations) → Webhook → Sentinel Data Connector → Log Analytics → Alerts/Analytics

Prerequisites​

  • Azure subscription with Microsoft Sentinel enabled
  • Log Analytics workspace created
  • Admin role in Noxys
  • Admin or Security Admin role in Azure

Step 1: Create Log Analytics Table​

Define where Noxys events are stored in Sentinel.

  1. Navigate to Log Analytics Workspace → Your workspace
  2. Go to Tables → Create → New custom table (DCR-based)
  3. Create table named: NoxysEvents_CL
  4. Define schema:
Column Name          | Type
--------------------|--------
TimeGenerated | datetime
EventType | string
TenantId | string
InteractionId | string
PlatformId | string
RiskScore | real
Classifications | dynamic
PolicyDecisions | dynamic
UserId | string
AlertSeverity | string
  1. Click Create

Step 2: Create Data Connector​

Enable webhook endpoint for Noxys to send events.

  1. In Sentinel → Data Connectors → Create new connector
  2. Choose API Polling or Webhook (REST API)
  3. Configure:

Name: Noxys AI Firewall

Webhook Configuration:

Request Method: POST
Content-Type: application/json
Authentication: Bearer token (generated in Noxys)
  1. Note the Webhook URL (provided by Sentinel)
  2. Create Data Connector Rule:
{
"name": "NoxysToSentinelRule",
"source": "NoxysEvents_CL",
"schedule": "PT1H"
}

Step 3: Configure Noxys Webhook​

Set up event forwarding from Noxys to Sentinel.

  1. In Noxys Admin Panel → Integrations → Webhooks
  2. Click Create Webhook
  3. Configure:

URL: https://your-workspace.eastus2.loganalytics.io/api/logs?api-version=2016-04-01 (from Sentinel)

Events: Select all or specific events:

  • interaction.created
  • interaction.policy_violated
  • interaction.high_risk
  • alert.severity_critical
  • alert.severity_high
  • policy.created
  • policy.updated

Headers:

Authorization: Bearer <workspace-key>
Content-Type: application/json

Description: "Microsoft Sentinel integration"

  1. Click Create
  2. Test webhook: Click Test
  3. Verify delivery in Deliveries tab

Step 4: Create Analytics Rules​

Define threat detection rules based on Noxys data.

Rule 1: Repeated Policy Violations​

In Sentinel:

  1. Analytics → Rule templates → Create rule → Scheduled query rule

  2. General:

    • Name: "High Volume Policy Violations from Noxys"
    • Severity: High
  3. Set rule logic:

NoxysEvents_CL
| where EventType == "interaction.policy_violated"
| summarize ViolationCount=count() by UserId, PlatformId, bin(TimeGenerated, 1h)
| where ViolationCount > 10
| project TimeGenerated, UserId, PlatformId, ViolationCount
  1. Query scheduling: Every 1 hour

  2. Alert details:

    • Entity mapping:
      • Account: UserId
    • Alert name: Multiple policy violations detected
  3. Create

Rule 2: High-Risk Interaction from Non-Enterprise Platform​

NoxysEvents_CL
| where EventType == "interaction.high_risk"
| where RiskScore > 0.8
| where PlatformId in ("perplexity", "deepseek", "grok")
| project TimeGenerated, UserId, PlatformId, RiskScore, Classifications
| take 100

Rule 3: Blocked Transactions on Unallowed Platforms​

NoxysEvents_CL
| where EventType == "interaction.policy_violated"
| where PolicyDecisions contains "block"
| where PlatformId not in ("chatgpt", "claude", "gemini")
| summarize BlockCount=count() by UserId, PlatformId, TimeGenerated

Step 5: Create Playbooks (Automation)​

Automatically respond to Noxys alerts.

In Sentinel → Automation → Create automation rule

Example: Create Ticket on Critical Violation​

Trigger: Alert created
Filter: Product = "Noxys"
Action: Create ticket in Jira/ServiceNow

Example: Notify Teams on High-Risk Activity​

Trigger: Alert created
Filter: Severity = High, AlertSeverity = critical
Action: Post message to Teams

Step 6: Create Dashboards​

Visualize Noxys data in Sentinel.

In Sentinel → Workbooks → + Add workbook

Add tiles:

  1. High-Risk Interactions (24h):
NoxysEvents_CL
| where EventType == "interaction.high_risk"
| where TimeGenerated > ago(24h)
| summarize Count=count() by PlatformId
  1. Policy Violations by Platform:
NoxysEvents_CL
| where EventType == "interaction.policy_violated"
| summarize ViolationCount=count() by PlatformId
| render barchart
  1. Risk Score Distribution:
NoxysEvents_CL
| where isnotnull(RiskScore)
| summarize AvgRisk=avg(RiskScore) by bin(TimeGenerated, 1h)
| render timechart
  1. Top Users with Violations:
NoxysEvents_CL
| where EventType == "interaction.policy_violated"
| summarize ViolationCount=count() by UserId
| top 10 by ViolationCount
| render barchart

Step 7: Verification​

Verify integration is working.

  1. In Noxys:

    • Create test interaction or trigger policy violation
    • Check webhook deliveries: Integrations → Webhooks → Deliveries
    • Verify status = 200
  2. In Sentinel:

    • Go to Logs → Run query:
    NoxysEvents_CL
    | where TimeGenerated > ago(1h)
    | take 10
    • Should see recent events
  3. Analytics Rules:

    • Trigger a policy violation
    • Check Incidents to see if rules fire

Troubleshooting​

Events Not Appearing in Sentinel​

  1. Check webhook is active: Admin Panel → Integrations → Webhooks → Status
  2. Verify webhook URL: Confirm it matches Sentinel's endpoint
  3. Check authentication: Verify workspace key is correct
  4. Review deliveries: Admin Panel → Webhooks → Deliveries
    • If status ≠ 200, check error message
  5. Firewall rules: Ensure Noxys IPs can reach Sentinel endpoint
  6. Wait for sync: LAW may take 1-5 minutes to appear

Analytics Rules Not Triggering​

  1. Check rule is Enabled
  2. Verify query is valid (test in Log Analytics)
  3. Check rule schedule is appropriate
  4. Review rule logic — verify it matches your data

High Latency​

  • Increase webhook batch size (if applicable)
  • Check network connectivity between Noxys and Azure
  • Review Log Analytics query performance

Cost Considerations​

Sentinel charges for:

  • Log ingestion: Per GB ingested
  • Analytics rules: Per rule execution
  • SOAR playbooks: Per action

Estimate: ~$10-50/month for typical deployment (varies by event volume)

Best Practices​

  1. Filter events wisely: Only send critical events to reduce costs

    • interaction.policy_violated, alert.severity_critical
    • Skip interaction.created (too verbose)
  2. Set appropriate thresholds:

    • Alert on 5+ violations/hour
    • High risk = RiskScore > 0.8
    • Critical severity = Block actions
  3. Use entity mapping: Link UserId and PlatformId for investigation

  4. Tune analytics rules: Test with historical data first

  5. Monitor alert fatigue: Review rules quarterly

Data Retention​

In Noxys: Default 90 days In Sentinel: Configurable (typically 30-730 days)

To extend retention in Noxys, contact support or upgrade plan.

Compliance Notes​

  • GDPR: Events contain hashed data only (no raw prompts)
  • SOC 2: Sentinel integration is SOC 2 Type II compliant
  • HIPAA: Requires HIPAA-enabled Log Analytics workspace

Sample Webhook Payload​

{
"id": "evt_abc123",
"timestamp": "2026-03-20T14:32:00Z",
"event_type": "interaction.policy_violated",
"tenant_id": "tenant-123",
"data": {
"id": "int_abc123",
"platform_id": "chatgpt",
"risk_score": 0.95,
"classifications": [
{
"type": "EMAIL",
"value": "alice@acme.fr",
"score": 0.98
}
],
"policy_decisions": [
{
"policy_id": "pol_abc123",
"policy_name": "Block PII on ChatGPT",
"action": "block"
}
]
}
}

Support & Resources​

Next Steps​

  • Create custom analytics rules for your use cases
  • Set up automated playbooks for incident response
  • Build dashboards for security team visibility
  • Export reports for compliance audit