---
license: mit
tags:
  - ueba
  - behavior-analytics
  - cybersecurity
  - synthetic
  - enterprise
  - iam
---

# Security Behavior Analytics — Synthetic IAM Behavior Dataset

100% synthetic anonymized enterprise IAM/access behavior events for UEBA cohort and peer analysis research.

## Schema

| Field | Type | Description |
|-------|------|-------------|
| `user_id` | string | Synthetic user identifier (USR-XXXXX) |
| `department` | string | Organizational department |
| `timestamp` | string | ISO 8601 event timestamp |
| `event_type` | string | login, file_download, resource_access, etc. |
| `device` | string | Windows Laptop, MacBook, VDI, Mobile |
| `location` | string | HQ, office, remote, VPN |
| `resource` | string | SharePoint, GitHub, SAP, AWS, etc. |
| `authentication_result` | string | success, failure, mfa_required, blocked |
| `file_download_size_mb` | float | Download volume in MB |
| `privilege_level` | string | Standard, Elevated, Admin |
| `session_duration_min` | float | Session length in minutes |
| `job_role` | string | Job title/role |
| `seniority` | string | Junior through Director |
| `work_mode` | string | Remote, On-site, Hybrid |

## Splits

| Split | Description |
|-------|-------------|
| `train.jsonl` | First 75% chronological |
| `validation.jsonl` | Next 15% |
| `test.jsonl` | Final 10% |
| `user_profiles.jsonl` | User metadata registry |

## Generation

Deterministic with `seed=42`. ~493K events, 320 users, 90-day window.

```bash
python scripts/export_publish_artifacts.py
```

## Privacy

No real user data. No PII. No production IAM exports.

## Related

- [Space Dashboard](https://huggingface.co/spaces/alirezaaminzadeh/security-behavior-analytics)
- [Model Parameters](https://huggingface.co/alirezaaminzadeh/security-behavior-analytics-models)