Skip to main content

Splunk Integration

Forward Noxys AI interaction events to Splunk via HTTP Event Collector (HEC) for centralized security monitoring and analysis.

Overview​

This integration enables:

  • Real-time event streaming from Noxys to Splunk
  • Custom searches and dashboards on AI interaction data
  • Automated alerting based on policy violations
  • Compliance reporting combining AI and security data

Data Flow:

Noxys Event → Webhook → Splunk HEC → Splunk Indexer → Splunk Search → Alerts & Dashboards

Prerequisites​

  • Splunk Enterprise or Cloud
  • HTTP Event Collector (HEC) enabled
  • Splunk admin access
  • Noxys admin role

Step 1: Create HEC Token​

Enable HTTP Event Collector and create a token.

  1. Splunk → Settings → Data Inputs → HTTP Event Collector
  2. Click New Token
  3. Configure:
    • Name: Noxys AI Firewall
    • Source name override: noxys
    • Default Index: main (or create custom index)
  4. Next
  5. Choose index:
    • Create custom: noxys_interactions
    • Or use existing: security
  6. Review settings
  7. Create Token
  8. Copy the token value (shown only once!)

Step 2: Verify HEC is Enabled​

Ensure HTTP Event Collector is globally enabled.

  1. Settings → Data Inputs → HTTP Event Collector
  2. In the toolbar, verify All Tokens is enabled (toggle should be ON)
  3. Note the HEC URL and Port:
    • Cloud: https://your-instance.splunkcloud.com:8088
    • Enterprise: https://your-splunk-server.com:8088

Step 3: Configure in Noxys​

Set up webhook forwarding to Splunk HEC.

  1. Noxys Admin Panel → Integrations → Webhooks
  2. Click Create Webhook
  3. Configure:
FieldValue
URLhttps://your-instance.splunkcloud.com:8088/services/collector
EventsSelect: interaction.policy_violated, interaction.high_risk, alert.severity_critical
HeadersAdd custom header: Authorization: Splunk <token-from-step-1>
Description"Splunk HEC Integration"
  1. Click Create

Step 4: Test Connection​

Verify events are flowing to Splunk.

  1. In Noxys: Click webhook → Test
  2. In Splunk:
    • Navigate to: Search & Reporting
    • Search: source="noxys"
    • Should see test event
  3. Create a test interaction in Noxys
  4. Should appear in Splunk within seconds

Step 5: Create Index (Optional)​

Create dedicated index for Noxys data.

  1. Settings → Indexes → New Index
  2. Configure:
    • Index name: noxys_interactions
    • Max KB: 500000 (reasonable default)
    • Max Raw Size MB: 1000
  3. Save
  4. Update HEC token to use this index (Step 1)

Step 6: Create Searches​

Build searches to analyze Noxys data.

Search 1: High-Risk Interactions (24h)​

source="noxys" event_type="interaction.high_risk"
| stats count by platform_id, risk_score
| sort - count

Search 2: Policy Violations by Platform​

source="noxys" event_type="interaction.policy_violated"
| stats count by data.platform_id as platform
| rename platform as Platform, count as "Violation Count"

Search 3: Top Users with Violations​

source="noxys" event_type="interaction.policy_violated"
| stats count by data.user_id
| rename data.user_id as User, count as Violations
| top 10 User

Search 4: Blocked Interactions Timeline​

source="noxys" event_type="interaction.policy_violated"
| search "policy_decisions{}.action"=block
| timechart count by data.platform_id

Step 7: Create Alerts​

Set up automated alerts for critical events.

Alert: Multiple Violations from One User​

  1. Search → Run search:
source="noxys" event_type="interaction.policy_violated"
| stats count as violation_count by data.user_id
| where violation_count > 10
  1. Save as Alert
  2. Configure:
    • Alert name: "High violation user detected"
    • Frequency: Every hour
    • Trigger condition: When violation_count > 10
  3. Add actions:
  4. Save

Alert: Blocked Access to Financial Data​

  1. Search:
source="noxys" event_type="interaction.policy_violated"
| search "data.classifications{}.type"=CREDIT_CARD OR "data.classifications{}.type"=IBAN
| search "policy_decisions{}.action"=block
  1. Save as alert, trigger immediately
  2. Add actions to escalate

Step 8: Build Dashboards​

Create visualization dashboards.

  1. Dashboards → Create Dashboard
  2. Add panels:

Panel 1: Interaction Volume (24h)

source="noxys" | timechart count

Panel 2: Policy Violations by Platform

source="noxys" event_type="interaction.policy_violated"
| stats count by data.platform_id
| chart count by data.platform_id

Panel 3: Risk Score Distribution

source="noxys"
| bucket data.risk_score range=0-0.2, 0.2-0.5, 0.5-0.8, 0.8-1.0
| stats count by data.risk_score
| chart count by data.risk_score

Panel 4: Top Classifications Detected

source="noxys"
| mvexpand data.classifications{}
| stats count by "data.classifications{}.type"
| top 10 "data.classifications{}.type"

Panel 5: Blocked Interactions

source="noxys" event_type="interaction.policy_violated"
| search "policy_decisions{}.action"=block
| stats count as blocked_count, latest(_time) as last_violation by data.user_id

Step 9: Verify Data​

Confirm events are indexed and searchable.

  1. Search & Reporting
  2. Search: source="noxys" earliest=-1h
  3. Should show recent events with fields:
    • event_type — interaction, alert, policy
    • data.platform_id — chatgpt, claude, etc.
    • data.risk_score — numeric risk
    • data.classifications{} — Risk Signals types detected
    • data.policy_decisions{} — Block/Coach/Log actions

Advanced: Log Forwarding​

Forward Splunk indexed data to other systems.

  1. Settings → Forwarding and receiving
  2. Configure destination:
    • Elasticsearch
    • S3
    • Syslog
    • HTTP endpoint
  3. Create forwarding rule:
    • Source index: noxys_interactions
    • Destination: Your endpoint
  4. Verify data flows through

Troubleshooting​

Events Not Appearing in Splunk​

  1. Verify HEC token:
    • Check token is active (not disabled)
    • Token may be single-use; test with Splunk verification
  2. Check webhook is active:
    • Noxys: Integrations → Webhooks → Status
  3. Review webhook deliveries:
    • Noxys: Webhooks → Deliveries
    • Look for HTTP status code
  4. Test HEC connectivity:
curl -k https://your-splunk.com:8088/services/collector \
-H "Authorization: Splunk <token>" \
-d '{"event":"test"}'
  1. Check Splunk HEC logs:
    • Splunk: Search & Reporting → source::splunkd token=<token>

Searches Return No Data​

  1. Verify event source matches:
    • Search: index=main source="noxys"
    • If no results, check HEC index configuration
  2. Verify event format:
    • Check raw events for JSON structure
    • Ensure required fields are present
  3. Check time range:
    • Default search: last 24 hours
    • Try: earliest=-7d for wider window

High Latency (Events delayed)​

  1. Check HEC input load:
    • Settings → Data Inputs → HTTP Event Collector → Monitor
    • Look for queued events
  2. Increase HEC throughput:
    • Settings → Data Inputs → HTTP Event Collector → Edit
    • Increase maxKBps if rate-limited
  3. Check network latency:
    • Ping Splunk HEC endpoint
    • Check firewall rules

Cost Optimization​

Splunk costs are based on daily indexing volume:

  • Noxys event size: ~2-5 KB per event
  • Daily volume estimate: 100 events/day = 200-500 KB
  • Monthly estimate: 6-15 MB
  • Typical Splunk cost: $100-300/month (depends on plan)

To reduce costs:

  1. Filter events before sending (send only critical ones)
  2. Increase webhook batch size
  3. Archive old logs to S3
  4. Use Splunk Cloud vs. Enterprise (lower cost)

Best Practices​

  1. Filter events wisely:

    • Send: policy_violated, high_risk, alert.critical
    • Skip: interaction.created (too verbose)
  2. Use custom index:

    • Separates Noxys data from security events
    • Easier to manage retention
  3. Implement field extraction:

    • Create props.conf for automatic field parsing
    • Extract platform_id, risk_score, user_id, action
  4. Monitor HEC health:

    • Create alert: "HEC connectivity lost"
    • Alert on missing events
  5. Archive data appropriately:

    • Hot: 7 days
    • Warm: 30 days
    • Cold: 90 days
    • Frozen: Delete or archive to S3

Splunk Configuration Files​

props.conf (Field Extraction)​

[source::noxys]
SHOULD_LINEMERGE = false
KV_MODE = json
TRANSFORMS = noxys_routing

[noxys_interactions]
SHOULD_LINEMERGE = false
KV_MODE = json

transforms.conf (Routing)​

[noxys_routing]
REGEX = .
DEST_KEY = _MetaData:Index
FORMAT = noxys_interactions

Support & Resources​