Skip to content

SIEM Integration Demo Script ​

Six demos. Sixteen minutes. Enterprise security telemetry from GhostAI to your SIEM.

This script walks through GhostAI’s SIEM integration: curated vendor presets with logos and setup guidance, webhook creation with sensitive-operation re-authentication, the machine-readable event catalog (including MITRE ATT&CK and detection rules), live test and validation against a real endpoint, operational health and delivery intelligence, and framework-aligned compliance reports.


Pre-Demo Checklist ​

  • [ ] GhostAI instance is running and reachable (UI and API)
  • [ ] Enterprise tier — SIEM integration is an Enterprise feature
  • [ ] SIEM integration enabled — org billing/feature flags include siem_integration (and backend has the SIEM dispatcher running if you expect live delivery)
  • [ ] Admin access — demo user can open Admin → SIEM (or equivalent) and call /api/admin/siem/* with a valid Bearer token
  • [ ] Re-authentication token ready — for create/delete credential flows, you can supply X-GhostAI-Reauth-Token (per org policy)
  • [ ] Target SIEM endpoint — Splunk HEC URL (or mock HTTP collector) that returns HTTP 2xx for a successful demo of Test and Validation Burst
  • [ ] Browser zoom and resolution comfortable for reading admin tables (1080p minimum recommended)
  • [ ] Optional: curl handy to show API parity with the UI

Quick API check ​

bash
curl -s -H "Authorization: Bearer $GHOSTAI_API_KEY" \
  https://your-host/api/admin/siem/presets | python3 -m json.tool

You should see a JSON array of presets with id, name, logo, and format. A 403 or feature error means Enterprise or siem_integration is not enabled for the org.


Duration: 2 minutes
Story: Show that GhostAI ships first-class integrations for the SIEMs enterprises already run—not a one-size-fits-all webhook.

Steps ​

  1. Open Admin → SIEM (or Integrations → SIEM) and select Presets / Browse presets.
  2. Scroll the gallery. Call out logos and default output format per vendor.

Presets to name explicitly (15 platforms + Custom) ​

Preset IDDisplay nameLogo assetDefault format
splunkSplunksplunk.svgsplunk_hec
sentinelMicrosoft Sentinelsentinel.svgsentinel
datadogDatadogdatadog.svgjson
elasticElasticelastic.svgjson
qradarIBM QRadarqradar.svgcef
chronicleGoogle Chroniclechronicle.svgjson
logscaleCrowdStrike Falcon LogScalelogscale.svgjson
cortex_xsiamPalo Alto Cortex XSIAMcortex_xsiam.svgjson
sumologicSumo Logicsumologic.svgjson
pantherPantherpanther.svgjson
exabeamExabeamexabeam.svgjson
logrhythmLogRhythmlogrhythm.svgcef
securonixSecuronixsecuronix.svgjson
devoDevodevo.svgjson
aws_security_lakeAWS Security Lakeaws_security_lake.svgocsf
customCustom / Othercustom.svgjson (configurable)
  1. Click Splunk (or any preset with a content pack). Mention downloadable content packs where available (e.g. Splunk add-on archive, QRadar content zip)—sourced from GET /api/admin/siem/content-packs/{preset_id} when packaged on the server.
  2. Open preset details and walk through setup steps (for Splunk: enable HEC, create token, index, allowlist outbound IPs, paste token).

Expected outcome ​

  • Audience sees 16 choices (15 named platforms + Custom), each with recognizable branding and the correct default serialization for that ecosystem.
  • Setup steps read like a runbook, not generic “paste URL.”

Talking points ​

"These aren’t vanity badges—we set the default format, auth header style, and batch defaults to what each vendor actually expects. Splunk defaults to HEC; QRadar leans CEF; Security Lake aligns with OCSF."

"Where we ship a content pack, your customer’s SOC can import dashboards and parsers in minutes instead of writing them from scratch."


Demo 2: Create an Integration ​

Duration: 3 minutes
Story: Stand up a Splunk HEC integration with operational tuning and least-privilege event selection.

Steps ​

  1. Click New integration → preset Splunk.
  2. Name: e.g. Prod – Splunk HEC.
  3. Delivery: Webhook (push) — events POST to your endpoint as they are batched.
  4. Endpoint URL: your HEC URL, e.g. https://splunk.example.com:8088/services/collector.
  5. Authentication: paste the HEC token. When the UI prompts for re-authentication, complete it so the client can send header X-GhostAI-Reauth-Token on the create request (required by the API for sensitive operations).
  6. Batching: set batch size (e.g. 100) and flush interval (e.g. 5 seconds)—explain throughput vs. latency.
  7. Event filters: configure JSON event filters on the integration (category lists, allow/deny event types, or other keys your deployment documents)—persisted for the dispatcher and UI.
  8. Format: confirm splunk_hec (preset default; overrides generic JSON for Splunk).
  9. Minimization policy: choose standard, minimal, or full—controls how much detail lands in details / actor fields for privacy and retention posture.
  10. Save.

Expected outcome ​

  • Integration appears in the list with masked credential tail (e.g. ••••abcd).
  • API equivalent: POST /api/admin/siem/integrations with Authorization: Bearer and X-GhostAI-Reauth-Token.

Talking points ​

"Creating an integration that stores credentials is a sensitive operation—we require a fresh re-auth signal via X-GhostAI-Reauth-Token, not just a long-lived session cookie."

"Batch size and flush interval let you match Splunk indexer capacity and licensing. Filters mean you don’t pay to store chatty events you’ll never alert on."


Demo 3: Event Catalog ​

Duration: 2 minutes
Story: Prove auditability: every streamable event is documented with samples, compliance tags, and MITRE mapping.

Steps ​

  1. Open Event catalog (or SIEM → Catalog).
  2. Browse by category—e.g. authentication, authorization, system—and note event counts per category.
  3. Open a high-value event such as authz.permission_denied or auth.login.
  4. On the detail page, show:
    • Description and field dictionary
    • Sample payload (structured details consistent with production shape)
    • MITRE ATT&CK technique ID when present (e.g. T1078 for valid accounts)
    • Compliance control references (e.g. SOC2, HIPAA, PCI-DSS strings)
  5. Open Detection rules (API: GET /api/admin/siem/catalog/detections) and tie one rule to the event you picked.

Expected outcome ​

  • Security leaders see coverage and mapping without reading source code.

Talking points ​

"This catalog is what your GRC team and MSSP use to answer ‘what can GhostAI prove happened?’—not a PDF appendix that drifts from reality."

"MITRE IDs are first-class. Your Sentinel or Chronicle teams can drop these straight into detection engineering backlogs."


Demo 4: Test and Validate ​

Duration: 2 minutes
Story: Close the loop: one test event, then a burst that exercises formatting and HTTP behavior.

Steps ​

  1. On the integration, click Send test event (POST /api/admin/siem/integrations/{id}/test).
  2. Show HTTP status and short response body (Splunk: success payload with code: 0).
  3. Run Validation burst (POST /api/admin/siem/integrations/{id}/validation-burst)—this sends many catalog-shaped events in one formatted payload; show events_sent in the response.
  4. Open Health (GET .../integrations/{id}/health): point to events per minute, success rate, and latency percentiles (p50 / p95 / p99) when stats exist.
  5. Optionally show Health score (GET .../integrations/{id}/health/score) as a single executive-friendly indicator.

Expected outcome ​

  • Test returns 2xx against the customer’s collector; burst proves scale and format correctness.

Talking points ​

"Test is a single canonical siem.test_sent event. Validation burst walks the catalog so your parser rules get exercised before go-live."

"Health rolls up delivery stats so your SOC manager sees trend, not just last success."


Demo 5: Health Monitoring ​

Duration: 3 minutes
Story: Operations: delivery history, errors with vendor-specific fixes, reconciliation, and gap detection.

Steps ​

  1. Delivery log: GET /api/admin/siem/integrations/{id}/deliveries — recent attempts, status, timestamps. Mention webhook log (GET .../webhook-log) for deeper HTTP-level traces when enabled.
  2. Errors: GET /api/admin/siem/integrations/{id}/errors — show failed deliveries with remediation text pulled from the preset’s error playbook (e.g. Splunk 401 → rotate HEC token).
  3. Reconciliation: GET /api/admin/siem/integrations/{id}/reconciliation?start_date=YYYY-MM-DD&end_date=YYYY-MM-DD — daily alignment of emitted vs. delivered counts for audit.
  4. Gap detection: GET /api/admin/siem/integrations/{id}/gaps — highlight sequence or time windows that need analyst follow-up.
  5. Tie back to preset remediation API: GET /api/admin/siem/presets/{preset_id}/error-remediation/{http_status} for scripted runbooks.

Expected outcome ​

  • Audience believes the integration is observable and recoverable, not “fire and forget.”

Talking points ​

"When Splunk returns 403, we don’t just show red—we explain index permissions and HEC enablement, because that advice is Splunk-specific."

"Reconciliation and gaps are how you answer auditors who ask whether any day’s log volume silently dropped."


Demo 6: Compliance Report ​

Duration: 2 minutes
Story: Map streamed telemetry to HIPAA, SOC 2, PCI-DSS, NIST 800-53, and GDPR control language.

Steps ​

  1. Open Compliance report (or call the API).
  2. Generate a report for soc2 (GET /api/admin/siem/compliance/report?framework=soc2).
  3. Show coverage: covered_event_types, total_event_types, coverage_percentage.
  4. Drill into event_types: each entry lists control IDs and severity.
  5. Repeat intent for hipaa, pci_dss, nist_800_53, gdpr (query parameter framework values match the API).
  6. Mention reconciliation block in the response when a date-bounded audit window is supplied (start_date, end_date query params).

Expected outcome ​

  • CISO sees framework vocabulary tied to concrete event types, not a generic “we log stuff” claim.

Talking points ​

"Coverage is computed from the same catalog your engineers use for parsers—so GRC and SecOps never argue about two different inventories."


Recovery Steps ​

403 reauth_required on create or delete ​

Symptom: API returns {"error":"reauth_required",...} when creating or deleting an integration.

Fix: Supply header X-GhostAI-Reauth-Token with a valid re-auth token from your IdP or admin policy. Retry the request.

Feature not available (non-Enterprise) ​

Symptom: 403 or middleware blocks /api/admin/siem/*.

Fix: Confirm org subscription is Enterprise and feature flag siem_integration is enabled. Restart backend after config changes if required.

Test or validation burst returns non-2xx ​

Symptom: Splunk/Datadog returns 401, 403, or 404.

Fix: Use preset remediation (UI errors panel or GET /api/admin/siem/presets/splunk/error-remediation/401). Typical fixes: rotate HEC token, fix index rights, correct URL/port, allowlist GhostAI egress IPs.

Pull API returns 401 ​

Symptom: GET /api/admin/siem/pull fails with invalid Bearer token.

Fix: Call POST /api/admin/siem/integrations/{id}/rotate-pull-token, store the returned pull_token in your secret manager, and use Authorization: Bearer <pull_token>. Pagination: cursor and limit (e.g. cursor=0, limit=10).

Empty health stats ​

Symptom: Latency and rates are null or zero right after creation.

Fix: Run test and validation burst, or wait for live traffic. Stats aggregate from delivery history over the configured lookback window.


Demo Flow Cheat Sheet ​

#DemoDurationPrimary artifactKey moment
1Presets gallery2 min16 presets + logosSplunk HEC + content pack story
2Create integration3 minSplunk webhookX-GhostAI-Reauth-Token + batch/filter tuning
3Event catalog2 minCatalog + detectionsMITRE + sample payload
4Test & validate2 minTest + burst + healthHTTP 2xx + events_sent
5Health monitoring3 minDeliveries + errors + gapsPreset-specific remediation
6Compliance report2 minframework=soc2 (etc.)Coverage % + control mapping
Total~14–16 min

Frequently Asked Questions (From Demos) ​

"Webhook vs pull—which should we use?"

Webhook (push): GhostAI POSTs batched events to your HTTPS endpoint; best when your SIEM exposes HEC, Log Analytics API, or a generic HTTP collector. Pull: your collector calls GET /api/admin/siem/pull with a Bearer pull token, using cursor and limit pagination to read queued events—ideal for strict ingress policies or batch jobs.

"Which output formats are supported?"

json (canonical envelope), cef, leef, splunk_hec, sentinel, and ocsf. Presets choose sensible defaults; Custom lets you pick json, cef, leef, or ocsf to match any HTTP receiver.

"What is the event envelope?"

Every event includes at minimum: id, timestamp, organization_id, event_type, category, severity, outcome, plus actor, resource, details, compliance, and optional sequence, prev_hash, source_table, signature. Schema: GET /api/admin/siem/schema.

"Do you support MSPs aggregating child tenants?"

Yes. Parent orgs with MSP aggregation enabled can use GET /api/admin/siem/msp/events with optional child_org_id and cursor pagination to pull telemetry across managed customers (requires backend MSP configuration).

"How do maintenance windows work?"

Schedule POST /api/admin/siem/integrations/{id}/maintenance with ISO start and end times—delivery pauses and metadata records the window. POST .../maintenance/cancel resumes normal flushing.


For step-by-step API exercises with curl, see the SIEM Integration Full Lab.

SOC 2 Type II Certified | BAA Available | HIPAA-Ready