QuailModel sample: QM Synthetic Network Intrusion Flows

README.md

@@ -9,116 +9,44 @@ tags:
 - synthetic
 - tabular
 - anomaly-detection
-pretty_name: QM Synthetic Network Intrusion Flows
-configs:
-- config_name: flows
-  data_files: flows_sample.csv
-  default: true
-- config_name: incidents
-  data_files: incidents.csv
+pretty_name: "QM Synthetic Network Intrusion Flows"
 ---
-# QuailModel Synthetic Network Intrusion Flows - Free Sample
+# QM Synthetic Network Intrusion Flows - Free Sample
 
-<!-- qm-buy:start -->
-**Buy the full commercial edition: $39 USD - launch price until 24 Oct 2026 (regular price $49 USD) -> [Polar checkout, instant download](https://buy.polar.sh/polar_cl_GGObtkVsr8IkPlgEdD2gD4M1kfp33AE84Jlr9360OY0)**  
-This free sample is non-commercial (CC BY-NC-SA 4.0). The paid full edition has a commercial licence.
-
-## Summary
 Synthetic enterprise network flow records with 7 labelled attack types and deliberate benign look-alikes, for intrusion-detection / SOC models; the paid full edition can be used commercially.
-- Best for: IDS / SOC model prototyping, feature engineering for per-host aggregates, detection-rule testing, teaching
-- Not for: claiming real-world detection rates or certifying a security product
 
-## Quick start
-```python
-from datasets import load_dataset
-ds = load_dataset("quailrobot/netflow-v1-sample", split="train")  # main table: flows_sample.csv
-# pandas alternative: pd.read_csv("hf://datasets/quailrobot/netflow-v1-sample/flows_sample.csv")
-```
+## What's inside
+- **This free sample:** 562,530 flows = first 3 days, all 7 attack types (flows_sample.csv) + the campaigns that start in that window
+- **Full commercial edition:** 4,291,839 flows, 1,500 hosts, 28 days, 168 attack campaigns, attack rate 0.0211 (zip 65 MB)
+- **Task:** tabular binary + multi-class classification (intrusion detection), anomaly detection, alert triage
+- **Labels:** label (0/1); attack_type = port_scan, ssh_brute_force, ddos, c2_beaconing, data_exfiltration, lateral_movement, dns_tunneling; campaign_id; incidents.csv (type, low_and_slow, start)
+- **Format:** full: Parquet (flows) + CSV (incidents); sample: CSV
 
-## Facts
-- Task: tabular binary + multi-class classification (intrusion detection), anomaly detection, alert triage
-- Labels: label (0/1); attack_type = port_scan, ssh_brute_force, ddos, c2_beaconing, data_exfiltration, lateral_movement, dns_tunneling; campaign_id; incidents.csv (type, low_and_slow, start)
-- Full edition size: 4,291,839 flows, 1,500 hosts, 28 days, 168 attack campaigns, attack rate 0.0211 (zip 65 MB)
-- Free sample size: 562,530 flows = first 3 days, all 7 attack types (flows_sample.csv) + the campaigns that start in that window
-- Format: full: Parquet (flows) + CSV (incidents); sample: CSV
-- What's included (full edition): flows.parquet, incidents.csv, stats.json, benchmark.json, bench_netflow.py, README.md, LICENSE.txt
-- Price: $39 USD, launch price until 24 Oct 2026 (regular price $49 USD); checkout may display the equivalent in your local currency
-- Buy URL (primary): https://buy.polar.sh/polar_cl_GGObtkVsr8IkPlgEdD2gD4M1kfp33AE84Jlr9360OY0
-- Licence (full edition): LicenseRef-QuailModel-Commercial (commercial use allowed, no resale of the data)
-- Licence (free sample): CC-BY-NC-SA-4.0
-- Validation: Our own test on the full edition (time split, last 11 days): gradient boosting on single-flow features ROC-AUC 0.937 / PR-AUC 0.260 (single flows are not enough); with 1-hour per-host aggregates 0.997 / 0.963; hard classes lateral_movement 0.020, data_exfiltration 0.456, ssh_brute_force 0.602 PR-AUC (benchmark.json + script included).
-- Data source: 100% synthetic, generated by QuailModel with AI assistance (generator code written with an AI model)
-- Catalog (all QuailModel datasets, catalog.json, llms.txt): https://quailrobot-quailmodel.static.hf.space
-- Last updated: 2026-10-10
+## Files
+flows (Parquet in the full edition, CSV in the sample), incidents.csv; the full edition adds stats.json,
+benchmark.json and bench_netflow.py.
+
+## How it was made
+Produced by QuailModel's own simulator, including legitimate activity alongside the labelled patterns. Labels are exact, because they are known by construction.
 
 ## Validation
 Our own test on the full edition (time split, last 11 days): gradient boosting on single-flow features ROC-AUC 0.937 / PR-AUC 0.260 (single flows are not enough); with 1-hour per-host aggregates 0.997 / 0.963; hard classes lateral_movement 0.020, data_exfiltration 0.456, ssh_brute_force 0.602 PR-AUC (benchmark.json + script included).
 
-## Price & licence
-- Full edition: $39 USD, launch price until 24 Oct 2026 (regular price $49 USD). One-time payment, instant download after checkout: https://buy.polar.sh/polar_cl_GGObtkVsr8IkPlgEdD2gD4M1kfp33AE84Jlr9360OY0
-- QuailModel Commercial Dataset Licence (SPDX: LicenseRef-QuailModel-Commercial): you may train, evaluate and ship models, including in commercial products. You may not resell or redistribute the dataset itself.
-- Free sample (this page): CC BY-NC-SA 4.0 - free for non-commercial use.
-
 ## Limitations
 - 100% synthetic: not validated against real network captures.
 - Flow records only - no packet payloads.
 - External IPs come only from RFC 5737 documentation ranges.
 - label, attack_type and campaign_id are labels - exclude them from model inputs.
 
-## Custom dataset of YOUR object ($249)
-**Need data of YOUR object? Custom synthetic dataset, $249 USD -> [order on Polar](https://buy.polar.sh/polar_cl_AQu6LzRtqiQmKPlgPt0zePBtvWjJI4vULeFMQ4NLHH9)**
-- Note: this is an image / object-detection service (not tabular data).
-- What you get: 2,000 labelled photoreal synthetic images (640x640 JPEG) of your own object or scenario (product, part, tool, drone, package, defect...), up to 3 classes, YOLO bounding boxes + data.yaml, train/val/test split, quality report
-- Licence: commercial use allowed
-- Price: $249 USD one-time; one round of adjustments included
-- Delivery: typically 3-5 business days after we receive your reference photos + rough dimensions
-- Optional sim-to-real test: send ~200 of your own labelled real images and we report how much the synthetic data improves a detector on them
-- Refund: full refund if we cannot deliver your request (14-day refund policy)
-- Limits: only objects you own or are allowed to use; no weapons or anything meant to harm people; no copied third-party 3D assets
-- Order URL: https://buy.polar.sh/polar_cl_AQu6LzRtqiQmKPlgPt0zePBtvWjJI4vULeFMQ4NLHH9
-- Details + contact: https://quailrobot-quailmodel.static.hf.space
-<!-- qm-buy:end -->
-
-- **What:** 562,530 labelled network flow records from a simulated 1,500-host enterprise network over 3 days,
-  with 19 attack campaigns of 7 types (port_scan, ssh_brute_force, ddos, c2_beaconing, data_exfiltration,
-  lateral_movement, dns_tunneling). Attack rate 0.0144 of flows.
-- **Full edition:** 4,291,839 flows, 1,500 hosts, 28 days, 168 campaigns (port_scan 36, ssh_brute_force 24,
-  data_exfiltration 24, c2_beaconing 23, ddos 22, dns_tunneling 21, lateral_movement 18).
-- **Columns:** flow_id, timestamp, src_ip, dst_ip, src_port, dst_port, protocol, bytes, packets, duration_s, tcp_flags;
-  labels: label (0/1), attack_type, campaign_id. incidents.csv lists each campaign (type, low_and_slow flag, start).
-- **No real hosts:** internal addresses are 10.0.0.0/16 (private); external addresses come ONLY from the RFC 5737
-  documentation ranges 192.0.2.0/24, 198.51.100.0/24, 203.0.113.0/24. No packet payloads, no real traffic, no PII.
-- **Use for:** intrusion-detection / SOC model prototyping, anomaly detection, alert triage, feature-engineering
-  teaching, detection-rule testing. **Not for:** claiming real-world detection rates.
+## Best for / not for
+Best for: IDS / SOC model prototyping, feature engineering for per-host aggregates, detection-rule testing, teaching. Not for: claiming real-world detection rates or certifying a security product.
 
-## Why single flows are not enough
-Many public intrusion datasets are near 100 % separable from single-flow features. Here every attack signature also
-occurs in benign traffic: all bare-SYN flows (internet background noise, a weekly internal vulnerability sweep incl.
-full-range sweeps, scans, SYN floods) share one size/duration distribution; SMTP deliveries from other mail servers,
-admin RDP / WinRM, Windows dynamic-RPC high ports, a partner SFTP server and a bastion with failed logins, legacy
-FTP / 8080 apps, flash crowds on the public web server, nightly backups, cloud-sync uploads, software-update check-ins
-every 15-60 min and chatty-DNS hosts with large responses. Packet counts follow bytes for every bulk transfer.
-About half of all attacks are low-and-slow; about a third of scans / brute-force runs come from a compromised internal
-host. We check this before release: a depth-3 decision tree on single-row features finds no near-pure (>= 99 %
-precision) rule covering >= 20 % of any attack type.
+## Credits
+No third-party data was used.
 
-## Benchmark (full edition, our test; time split: train first 17 days, test last 11)
-HistGradientBoosting (200 iterations), test set 1,598,989 flows (attack rate 0.0229):
-- single-flow features only: ROC-AUC 0.937, PR-AUC 0.260 - single flows are NOT enough
-- + simple per-source / per-destination 1-hour aggregates: ROC-AUC 0.997, PR-AUC 0.963
-- per-type PR-AUC with aggregates: ddos 0.988, c2_beaconing 0.901, dns_tunneling 0.890, port_scan 0.866,
-  ssh_brute_force 0.602, data_exfiltration 0.456, lateral_movement 0.020 - lateral movement and exfiltration leave
-  real room for better models (graph / sequence / per-host baselines). The overall score is dominated by
-  high-volume attacks; report per-type metrics.
-(benchmark.json + bench_netflow.py are included in the full edition.)
-
-## Files
-flows (Parquet in the full edition, CSV in the sample), incidents.csv; the full edition adds stats.json,
-benchmark.json and bench_netflow.py.
+## Licence
+Sample edition: CC BY-NC-SA 4.0 (non-commercial). The full commercial edition is sold by QuailModel (see the buy link).
 
 ## Disclosure
-Generated by QuailModel with AI assistance (simulator code written with an AI model). All traffic is synthetic;
-validate on your own network data before production use.
-
-## Licence
-Sample edition: CC BY-NC-SA 4.0 (non-commercial). The full commercial edition is sold by QuailModel.
+Generated by QuailModel with AI assistance (generator code written with an AI model); all data is synthetic /
+computer-generated. Validate on your own real data before production use.