cyb008-sample / README.md
pradeep-xpert's picture
Update README.md
3ad06d2 verified
---
license: cc-by-nc-4.0
task_categories:
- tabular-classification
- time-series-forecasting
tags:
- cybersecurity
- soc-operations
- alert-triage
- mitre-attack
- soar
- siem
- synthetic-data
- incident-response
- analyst-fatigue
- false-positive-reduction
pretty_name: CYB008 Synthetic SOC Alert Dataset (Sample)
size_categories:
- 10K<n<100K
---
# CYB008 — Synthetic SOC Alert Dataset (Sample)
**XpertSystems.ai Synthetic Data Platform · SKU: CYB008-SAMPLE · Version 1.0.0**
This is a **free preview** of the full **CYB008 — Synthetic SOC Alert
Dataset** product. It contains roughly **~10% of the full dataset** at
identical schema, MITRE ATT&CK tactic coverage, and statistical fingerprint,
so you can evaluate fit before licensing the full product.
> 🤖 **Trained baseline available:**
> [**xpertsystems/cyb008-baseline-classifier**](https://huggingface.co/xpertsystems/cyb008-baseline-classifier)
> — XGBoost + PyTorch MLP for **5-class SOC alert triage outcome
> classification** (the README's first headline use case), stratified
> split, multi-seed evaluation (ROC-AUC 0.955 ± 0.003). **Includes a
> structural-leakage diagnostic** documenting three oracle columns
> dropped from the feature set, and a separate unlearnable-target
> finding for MITRE ATT&CK tactic classification. Buyers planning
> SOC ML work should read the diagnostic first.
| File | Rows (sample) | Rows (full) | Description |
|----------------------------|---------------|---------------|----------------------------------------------|
| `soc_topology.csv` | ~25 | ~2,400 | SOC / analyst registry |
| `incident_summary.csv` | ~589 | ~4,800 | Per-incident aggregate outcomes |
| `alert_events.csv` | ~55,298 | ~48,000 | Discrete alert event log |
| `soc_alerts.csv` | ~9,200 | ~280,000 | Per-alert records (primary file) |
## Dataset Summary
CYB008 simulates end-to-end Security Operations Centre (SOC) alert
lifecycles across enterprise detection environments, with:
- **Full MITRE ATT&CK tactic coverage** — alerts mapped to all 14
Enterprise tactics from reconnaissance through impact
- **Alert severity distribution** — info / low / medium / high / critical
/ false_positive, with calibrated 45% false-positive baseline
- **SOC analyst tier modeling** — tier_1 / tier_2 / tier_3 / SOC manager
with differentiated MTTR by experience level
- **SOAR automation** — playbook trigger probability, auto-resolution
rate, automation coverage by alert type
- **Alert storm events** — Poisson-distributed alert bursts (2.5×–6×
amplification) simulating coordinated attacks or system failures
- **Analyst fatigue modeling** — utilization-driven burnout with MTTR
degradation past fatigue threshold (0.82)
- **Kill-chain correlated incidents** — alerts grouped into multi-stage
incidents when ≥3 ATT&CK tactics observed
- **SLA tracking** — severity-dependent SLA thresholds (critical 60min,
high 240min, medium 480min, low 1440min)
- **Detection source mix** — EDR, SIEM, NDR, IDS, UEBA, CASB, deception,
threat intel feeds
- **Rule drift modeling** — periodic rule-noise amplification simulating
detection-engineering signal decay
## Trained Baseline Available
A working baseline classifier trained on this sample is published at
**[xpertsystems/cyb008-baseline-classifier](https://huggingface.co/xpertsystems/cyb008-baseline-classifier)**.
| Component | Detail |
|---|---|
| Primary task | **5-class `resolution_outcome` classification** (SOC alert triage — the README's first headline use case) |
| Diagnostic | Structural-leakage audit (3 oracle columns dropped) + unlearnable-target finding for `mitre_tactic` |
| Models | XGBoost (`model_xgb.json`) + PyTorch MLP (`model_mlp.safetensors`) |
| Features | 53 (after one-hot encoding); pipeline included as `feature_engineering.py` |
| Split | **Stratified random** — no natural row-level group key (25 analysts, 5 SOCs, only 9% of alerts link to an incident) |
| Validation | Single seed + multi-seed aggregate across 10 seeds |
| Demo | `inference_example.ipynb` — end-to-end copy-paste |
| Headline metrics | XGBoost: accuracy 0.777 ± 0.007, macro ROC-AUC 0.955 ± 0.003 (multi-seed); MLP slightly outperforms |
**Important findings for buyers planning SOC ML work:**
1. **Three structural oracles in the data** (`alert_lifecycle_phase`,
`automation_resolved`, `escalation_flag`) deterministically encode
the `resolution_outcome` label. With these columns present, a
plain XGBoost achieves 100% accuracy. The baseline excludes them
to demonstrate honest learning — and the documented honest result
(acc 0.78, AUC 0.96) is genuinely useful.
2. **MITRE ATT&CK tactic classification is NOT learnable on this
sample.** The README lists tactic classification as a top use case,
but feature distributions are nearly identical across all 12
tactics. A trained model performs *below* majority baseline
(acc 0.08 vs 0.14). The baseline model card documents this
explicitly with a recommendation to the dataset author.
3. **SLA breach prediction is also not learnable** (acc 0.68 vs
majority 0.82). Documented as out-of-scope.
See the model card and `leakage_diagnostic.json` for the full audit
and our recommendations to make these tasks viable in the next
dataset version.
## Calibrated Benchmark Targets
The full product is calibrated to 12 benchmark validation tests drawn from
authoritative SOC operations research (SANS SOC Survey, IBM Cost of Data
Breach, Mandiant M-Trends, Forrester Wave SOAR, Gartner SIEM Magic
Quadrant, SOC.OS, CrowdStrike, Splunk State of Security, Verizon DBIR).
Sample benchmark results:
| Test | Target | Observed | Verdict |
|------|--------|----------|---------|
| false_positive_rate | 0.4500 | 0.4518 | ✓ PASS |
| mttd_minutes_mean | 132.0 | 137.1 | ✓ PASS |
| mttr_minutes_mean | 480.0 | 494.9 | ✓ PASS |
| escalation_rate | 0.2200 | 0.2038 | ✓ PASS |
| auto_resolution_rate | 0.3100 | 0.2872 | ✓ PASS |
| alert_volume_rate | 0.1650 | 0.1840 | ✓ PASS |
| analyst_fatigue_score | 0.6400 | 0.6457 | ✓ PASS |
| soar_playbook_execution_rate | 0.4200 | 0.4223 | ✓ PASS |
| incident_declaration_rate | 0.0850 | 0.0640 | ✓ PASS |
| true_positive_rate | 0.3800 | 0.3442 | ✓ PASS |
| kill_chain_completion_rate | 0.1450 | 0.1290 | ✓ PASS |
| sla_breach_rate | 0.1800 | 0.1775 | ✓ PASS |
*Note: every CYB008 benchmark is directly parametrised by the generator
(e.g. `soar_trigger_prob=0.42` produces `soar_playbook_execution_rate=0.42`).
Calibration is precise even at sample scale. The full product produces the
same calibration across 28× more data.*
## Schema Highlights
### `soc_alerts.csv` (primary file)
| Column | Type | Description |
|---------------------------------|---------|----------------------------------------------|
| alert_id | string | Unique alert identifier |
| incident_id | string | Parent incident FK (nullable) |
| soc_id | string | SOC environment FK |
| analyst_id | string | Assigned analyst FK |
| alert_timestamp | string | ISO timestamp |
| alert_severity | string | info / low / medium / high / critical / false_positive |
| mitre_tactic | string | 1 of 14 ATT&CK tactics |
| mitre_technique_id | string | T-number (e.g. T1059.001) |
| detection_source | string | edr / siem / ndr / ids / ueba / casb / etc. |
| triage_score | float | Initial triage priority (0–1) |
| enrichment_score | float | Threat-intel enrichment quality (0–1) |
| escalation_flag | int | Boolean — escalated to tier 2/3 |
| automation_resolved | int | Boolean — SOAR auto-resolved |
| soar_playbook_triggered | int | Boolean — SOAR playbook executed |
| mttd_minutes | float | Mean time to detect |
| mttr_minutes | float | Mean time to respond |
| sla_breached_flag | int | Boolean — SLA breached |
| resolution_outcome | string | true_positive / false_positive / duplicate / suppressed |
| analyst_tier | string | tier_1 / tier_2 / tier_3 / manager |
| storm_event_flag | int | Boolean — part of alert storm |
| kill_chain_stage | int | Position in kill chain (0 if standalone) |
### `incident_summary.csv` (per-incident outcome)
| Column | Type | Description |
|---------------------------------|---------|----------------------------------------------|
| incident_id | string | Identifier |
| soc_id, analyst_id | string | Identifiers |
| n_alerts_correlated | int | Alerts grouped into this incident |
| kill_chain_stages_observed | int | Distinct ATT&CK tactics in chain |
| incident_severity | string | Composite severity |
| incident_resolution_outcome | string | true_positive / false_positive / partial |
| analyst_fatigue_score | float | Avg fatigue during incident (0–1) |
| incident_duration_minutes | float | End-to-end response time |
See `alert_events.csv` and `soc_topology.csv` for the discrete event log
and SOC registry schemas respectively.
## Suggested Use Cases
- Training **alert triage** models — predict TP vs FP, or full 5-class
resolution outcome (the baseline ships this) —
[worked example available](https://huggingface.co/xpertsystems/cyb008-baseline-classifier)
- **MITRE ATT&CK tactic classification** from alert features (see baseline diagnostic — not learnable on this sample)
- **SOAR playbook recommendation** — predict which alerts benefit from
automation
- **Alert prioritization** — calibrate triage scores against ground-truth
outcomes
- **Analyst fatigue forecasting** — predict burnout from shift-level
workload
- **Kill-chain detection** — group related alerts into multi-stage
incidents
- **SLA breach prediction** — early-warning systems (see baseline diagnostic — not learnable on this sample)
- **Alert storm detection** — distinguish coordinated bursts from baseline
volume
- **False positive reduction** modeling — reduce 45% FP rate
- **Detection rule tuning** — identify rules with high noise factor
## Loading the Data
```python
import pandas as pd
alerts = pd.read_csv("soc_alerts.csv")
incidents = pd.read_csv("incident_summary.csv")
events = pd.read_csv("alert_events.csv")
topology = pd.read_csv("soc_topology.csv")
# Join alerts with analyst context
enriched = alerts.merge(topology, on=["soc_id", "analyst_id"], how="left",
suffixes=("", "_analyst"))
# 5-class triage outcome target (the README's first headline use case)
y_outcome = alerts["resolution_outcome"]
# Binary true-positive collapse (for binary triage)
y_tp = alerts["resolution_outcome"].isin([
"true_positive_remediated", "true_positive_escalated",
]).astype(int)
# Multi-class ATT&CK tactic classification target — see leakage diagnostic
y_tactic = alerts["mitre_tactic"]
# Binary SLA breach prediction target — see leakage diagnostic
y_sla = alerts["sla_breached_flag"]
```
For a worked end-to-end example with 5-class triage classification,
stratified splitting, and feature engineering, see the inference notebook
in the [baseline classifier repo](https://huggingface.co/xpertsystems/cyb008-baseline-classifier/blob/main/inference_example.ipynb).
## License
This **sample** is released under **CC-BY-NC-4.0** (free for non-commercial
research and evaluation). The **full production dataset** is licensed
commercially — contact XpertSystems.ai for licensing terms.
## Full Product
The full CYB008 dataset includes **~335,000 rows** across all four files,
with calibrated benchmark validation against 12 metrics drawn from
authoritative SOC operations and threat intelligence sources.
📧 **pradeep@xpertsystems.ai**
🌐 **https://xpertsystems.ai**
## Citation
```bibtex
@dataset{xpertsystems_cyb008_sample_2026,
title = {CYB008: Synthetic SOC Alert Dataset (Sample)},
author = {XpertSystems.ai},
year = {2026},
url = {https://huggingface.co/datasets/xpertsystems/cyb008-sample}
}
```
## Generation Details
- Generator version : 1.2.0
- Random seed : 42
- Generated : 2026-05-16 14:24:43 UTC
- Alert lifecycle : Multi-phase state machine with SOAR / fatigue / storm
- Overall benchmark : 100.0 / 100 (grade A+)