security-generate-security-sample-data（security 生成安全样本数据）

Generate Security Sample Data

Generate ECS-compliant security events, multi-step attack scenarios, and synthetic alert documents that populate Elastic Security dashboards, the Alerts tab, and Attack Discovery.

Quick start

For a zero-friction experience that generates everything and opens Kibana:

node skills/security/generate-security-sample-data/scripts/demo-walkthrough.js

Workflow

- [ ] Step 1: Set environment variables
- [ ] Step 2: Generate sample data
- [ ] Step 3: Explore in Kibana
- [ ] Step 4: Clean up when done

Step 1: Set environment variables

export ELASTICSEARCH_URL="https://your-project.es.region.aws.elastic.cloud"
export ELASTICSEARCH_USERNAME="admin"
export ELASTICSEARCH_PASSWORD="your-password"
export KIBANA_URL="https://your-project.kb.region.aws.elastic.cloud"

Step 2: Generate sample data

Generate everything at once

node skills/security/generate-security-sample-data/scripts/sample-data.js \
  system endpoint okta aws windows --scenarios --alerts

Generate only events

node skills/security/generate-security-sample-data/scripts/sample-data.js \
  system endpoint --count 100

Generate only attack scenarios

node skills/security/generate-security-sample-data/scripts/sample-data.js --scenarios

Generate only synthetic alerts

node skills/security/generate-security-sample-data/scripts/sample-data.js --alerts

Step 3: Explore in Kibana

After generating data, direct the user to these pages:

• Security > Alerts — synthetic alerts with MITRE ATT&CK mappings
• Security > Attack Discovery — requires an LLM connector to analyze alerts
• Security > Hosts — host activity from sample events
• Security > Overview — summary of all security data
• Discover — raw events across all data streams

Step 4: Clean up when done

node skills/security/generate-security-sample-data/scripts/sample-data.js --cleanup

What gets generated

Sample data spans 5 packages (system, endpoint, windows, aws, okta) and 4 focused attack scenarios covering the most common demo themes: Windows credential theft, AWS cloud privilege escalation, Okta identity takeover, and a full ransomware kill chain. Synthetic alert documents are indexed into .alerts-security.alerts-default with MITRE ATT&CK mappings, severity levels, and risk scores.

All events use RFC 5737 / RFC 2606 safe addresses. For full tables of packages, scenarios, and alerts see references/sample-data-reference.md.

Continuous mode

Stream events to simulate a live environment:

node skills/security/generate-security-sample-data/scripts/sample-data.js \
  --continuous --interval 15

Every 5th batch includes an attack scenario; every 10th batch adds synthetic alerts. Press Ctrl+C to stop.

Tool reference

sample-data.js

demo-walkthrough.js

Zero-friction runner that generates everything and opens Kibana.

Examples

Quick demo for a stakeholder

"Set up a demo environment so I can show Attack Discovery to my VP."

node skills/security/generate-security-sample-data/scripts/demo-walkthrough.js

Targeted scenario testing

"Generate only the ransomware attack chain to test our detection rules."

node skills/security/generate-security-sample-data/scripts/sample-data.js \
  --scenario ransomwareChain --alerts

Simulating a live SOC

"Keep generating events so the dashboards stay active during the demo."

node skills/security/generate-security-sample-data/scripts/demo-walkthrough.js --continuous

Cleaning up after a demo

"Remove all sample data from my project."

node skills/security/generate-security-sample-data/scripts/sample-data.js --cleanup

Guidelines

• All generated documents are tagged with tags: ["elastic-security-sample-data"] for safe cleanup. The cleanup command only deletes documents with this marker.
• If marker fields are not indexed in a data stream, cleanup falls back to scanning _source.tags for matching sample documents from the last 14 days.
• Synthetic alerts are indexed directly into .alerts-security.alerts-default — they do not require detection rules to be installed or enabled.
• Attack Discovery requires an LLM connector (OpenAI, Anthropic, Google Gemini, or similar) configured in Kibana under Stack Management > Connectors. The "Complete" project tier unlocks the feature, but the connector must be set up separately.
• Use the case-management skill for creating investigation cases from alerts.

Production use

• Do not run against production clusters unless you intend to inject synthetic data alongside real alerts. Sample events and alerts are tagged for cleanup but will appear in dashboards, the Alerts tab, and Attack Discovery alongside real data.
• All write operations (generate, --cleanup, --continuous) prompt for confirmation. Pass --yes or -y to skip when called by an agent.
• --cleanup runs deleteByQuery across all sample data indices — verify environment variables point to the intended cluster before running.
• --continuous mode indexes events indefinitely until manually stopped with Ctrl+C.

Environment variables

*Either API key or username/password is required for Elasticsearch.